Western Michigan University Western Michigan University

ScholarWorks at WMU ScholarWorks at WMU

Dissertations Graduate College

12-2013

Foreign Aid Effectiveness: Three Essays on Aid-For-Trade and Foreign Aid Effectiveness: Three Essays on Aid-For-Trade and

Export Performance of Developing Countries Export Performance of Developing Countries

Shanka Prasad Ghimire Western Michigan University, shanmire@gmail.com

Follow this and additional works at: https://scholarworks.wmich.edu/dissertations

Part of the International Economics Commons

Recommended Citation Recommended Citation Ghimire, Shanka Prasad, "Foreign Aid Effectiveness: Three Essays on Aid-For-Trade and Export Performance of Developing Countries" (2013). Dissertations. 204. https://scholarworks.wmich.edu/dissertations/204

This Dissertation-Open Access is brought to you for free and open access by the Graduate College at ScholarWorks at WMU. It has been accepted for inclusion in Dissertations by an authorized administrator of ScholarWorks at WMU. For more information, please contact wmu-scholarworks@wmich.edu.

FOREIGN AID EFFECTIVENESS: THREE ESSAYS ON AID-FOR-TRADE AND EXPORT PERFORMANCE OF DEVELOPING COUNTRIES

by

Shankar Prasad Ghimire

A dissertation submitted to the Graduate College in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Economics Western Michigan University

December 2013

Doctoral Committee:

Eskander Alvi, Ph.D., Co-Chair Debasri Mukherjee, Ph.D., Co-Chair Mahendra Lawoti, Ph.D.

FOREIGN AID EFFECTIVENESS: THREE ESSAYS ON AID-FOR-TRADE AND EXPORT PERFORMANCE OF DEVELOPING COUNTRIES

Shankar Prasad Ghimire, Ph.D.

Western Michigan University, 2013

This dissertation examines the effectiveness of foreign aid in enhancing

international trade in developing countries. In that light, it presents three essays

that focus on foreign aid targeted towards improving trade capacity of developing

countries (Aid-for-Trade or AfT) and analyzes its impact on export performance of

121 AfT-recipient countries over a period of 16 years [1995-2010].

The first essay examines whether aggregate AfT helps aid recipients improve

their aggregate export performance. The analysis using System-GMM shows a

positive and significant impact of AfT on the level and growth of exports as well as

export volume relative to GDP. These findings indicate that AfT can be effective in

stimulating overall exports in aid receiving countries. However, such targeted aid is

found to exhibit diminishing returns, suggesting AfT’s limited role in the

development of trade capacity.

The second essay concentrates on compilation of sector-wise disaggregated

aid-for-trade (SAfT) for agriculture, manufacturing, and service sectors and examines

the impact of SAfT on corresponding sectoral export levels. Since exports in one

sector can be correlated with exports in other sectors within a country, a Seemingly

Unrelated Regression (SUR) framework is used to capture the interdependence

among various sectors in an explicit way, producing efficient empirical estimates.

The results show, in most cases, that SAfT is effective in enhancing corresponding

sectoral exports.

The third essay examines the role of AfT on reducing export variability. The

data are purely cross-sectional with only 121 available observations, but there are

many likely regressors that impact export variance, creating an interesting model

selection problem. This study uses Least Angle Regression (LARS) as the model

selection method to assess the effectiveness of AfT to pinpoint the set of predictors

that are statistically robust and have strong predictive power, finding that AfT is one

such variable.

Overall results indicate that AfT initiatives favorably impact both the level

and variability measures of exports. This is strong evidence in support of such

targeted aid, offering a potent channel of economic expansion and growth in

developing countries.

ACKNOWLEDGMENTS

This dissertation is a product of valuable guidance, strong inspiration, and

continuous support I received from my professors, family members, and friends

around me. I am thankful to everyone in the Department of Economics and the

Graduate College at Western Michigan University for providing an avenue for my

professional development. My special appreciations go to Dr. Eskander Alvi for

explaining complex economic theories in a simple and intuitive manner. It is his

enthusiasm about foreign aid and economic growth of developing countries that

guided me to dedicate my dissertation on the effectiveness of Aid-for-Trade. In

addition, the econometric models used in this dissertation would not have made

much sense without thorough guidance from Dr. Debasri Mukherjee. Also, I am

grateful to Dr. Mahendra Lawoti for serving on my committee and for providing

valuable suggestions. Finally, I would like to thank Dr. Donald Meyer and Dr. Susan

Pozo for helping me in my transition from a graduate student to a professional

educator.

Shankar Prasad Ghimire

ii

TABLE OF CONTENTS

ACKNOWLEDGMENTS ............................................................................................. ii

CHAPTER

1. INTRODUCTION .......................................................................................... 1

2. AID FOR TRADE AND AGGREGATE EXPORT PERFORMANCE OFDEVELOPING COUNTRIES: A DYNAMIC PANEL ESTIMATION .................... 12

2.1. Introduction …………………..………………………………………………….. ..... 12

2.2. Conceptual Framework for Analyzing Aid-for-Trade …… ........... 15

2.3. Data and Empirical Analysis ………………………………………………. ...... 19

2.3.1. Measuring Aid-for-Trade ……………………………………… ......... 20

2.3.2. Measuring Export Performance ……………………………. ........ 22

2.3.3. Model Specification and Estimation Technique ……… ....... 23

2.4. Empirical Results …………………………………………………………… .......... 25

2.4.1. Explaining Export Levels ………….………………………….. .......... 25

2.4.2. Explaining Diminishing Returns to Aid-for-Trade …. .......... 29

2.4.3. Explaining Export Growth and Change in Export-GDP Ratio ……………………………………………………………………………….. ....... 31

2.5. Conclusions with Discussions ...……………………………. ................... 32

3. SECTORAL AID AND SECTORAL EXPORTS: A SEEMINGLY UNRELATEDREGRESSION ANALYSIS .............................................................................. 35

3.1. Introduction …………………..…………………………………………………… .... 35

3.2. Data and Measuring Sectoral Aid-for-Trade …………………… ......... 37

iii

Table of Contents – continued

3.3. Model Specification and Estimation Technique ……………… ......... 38

3.4. Empirical Results …………………………………………………………… .......... 40

3.5. Conclusions ...........................….…………………………………….. .......... 42

4. EFFECTIVENESS OF AID-FOR-TRADE ON REDUCING EXPORTVARIABILITY: SEARCHING FOR BEST PREDICTORS USING LARS ................ 44

4.1. Introduction …………………..…………………………………………………. ...... 44

4.2. Review of Literature ………………………………….…………………… ......... 47

4.3. Data and Model ……………………………………………..……………….. ....... 50

4.4. Empirical Results …………………………………………………………… .......... 54

4.5. Conclusions with Discussions ………………………………………...... ....... 57

5. CONCLUSIONS ........................................................................................... 60

REFERENCES ........................................................................................................... 66

APPENDICES A. OFFICIAL DEVELOPMENT ASSISTANCE: FOREIGN AID CATEGORIES ........... 72

B. SUMMARY STATISTICS ................................................................................ 73

C. LIST OF COUNTRIES INCLUDED IN THE STUDY ............................................ 74

D. SUPPLEMENTAL TABLES TO CHAPTER 2 ..................................................... 75

E. SUPPLEMENTAL TABLES TO CHAPTER 3 ..................................................... 81

F. SUPPLEMENTAL TABLES TO CHAPTER 4 ..................................................... 84

iv

CHAPTER 1

INTRODUCTION

The practice of giving foreign aid for development projects (development

assistance) has existed for more than half a century, along with much passion,

debate and scrutiny about its effectiveness. In the immediate period after the

Second World War, when the United States demonstrated the value of providing

financial assistance to rebuild a war-torn Europe, a notable increase in development

assistance emerged from other rich countries and financial communities toward

areas of the world in need. Since then, developed countries have been providing

assistance to less developed countries in different forms and for different purposes.

Poverty alleviation and institutional quality improvement, two prominent objectives,

are just a few in the wide range of categories today that aim to promote economic

wellbeing in recipient countries.

While allocation of such financial assistance may be dependent on political

and other strategic interests of donors, it is also anticipated to equally meet the

economic needs of the recipients. The tradition of assisting countries through aid is

generally accepted as indisputable given the mostly principled motivations and

expectations. However, foreign aid effectiveness is intensely and passionately

debated by economists, politicians, financiers and others. Arguing for and against

1

foreign aid continues to polarize researchers as well, producing sizeable literatures

and findings that are contrary to each other.

A survey of the literature shows that discussions of aid effectiveness were

divided into two groups until the mid-1990s; simply put, the perception was that

either aid works or it does not. However, the publication of ‘Assessing Aid: What

Works and What Does Not and Why’ in 1998 by the World Bank caught the attention

of policy makers involved in international development about the role of economic

policies in the recipient countries (McGillivray et al., 2006). The subsequent seminal

publication by Burnside and Dollar in 2000 added another dimension to this debate

by claiming that foreign aid is effective only in a good policy environment, which

revived the discussion on foreign aid effectiveness relating to fiscal, monetary, and

trade policies. Since this revival, many other studies reported similar findings

supporting this claim (Collier and Dehn, 2001; Collier and Dollar, 2002; Burnside and

Dollar, 2004). Nonetheless, these claims were contested for not being robust

enough to a change in the sample of countries or in the set of regressors used in the

analysis (Dalgaard and Hansen, 2001; Islam, 2002; Easterly et al. 2004).

Even with years of research accumulating on foreign aid effectiveness, the

controversy still continues to grow. A group of economists, represented forcefully by

Jeffery Sachs, explicitly advocates for increased aid assistance in support of the “big

push” theory, while another group of economists, prominently represented by

2

William Easterly, criticizes aid initiatives as being “white man’s burden”.1 To confuse

the argument further, a recent survey of the literature on the effectiveness of

foreign aid by Doucouliagos and Paldam (2009) indicates that the majority of

development assistance in the last forty years has not been productive. It is

therefore natural to question why so many of these studies keep showing foreign

aid as not being able to meet its economic expectations when there also exists

evidence showing the opposite. And, most importantly, the concern raises the

question whether it might be the continued use of a standard approach used in

measuring foreign aid that has influenced previous conclusions.

This dissertation argues that the ineffectiveness of foreign aid shown in the

past studies may be due to the fact that foreign aid has only been examined on the

basis of a summative view with the assumptions as well as expectations that every

foreign aid category produces the same results. The majority of these studies

analyze the impact of total Official Development Assistance (ODA) on overall

economic growth, which ignores the vagaries within categories and, in fact, that aid

in one category may be more effective than aid in another. I argue that not all of the

“general aid” (ODA) is streamlined to projects that influence economic growth

1 Jeffery Sachs and William Easterly are prominent representative figures in this debate; even in their recent works, they argue for and against, respectively, current trend of foreign aid allocation and its effectiveness. Sachs (2005) explains the role of foreign aid in alleviating poverty from developing countries and claims that ‘big push’, which involves injecting substantial amount of financial resources instead of taking one step at a time, is the only solution to uplift poverty stricken Africa. Easterly (2006) on the other hand criticizes foreign aid for not having a significant impact on economic growth of the recipients; instead, he argues, it is a burden to the West and that aid is promoting corruption and institutional failure by financing the interests of the elites in aid receiving countries.

3

directly but may have favorable impacts indirectly. These claims, therefore, highlight

a need to analyze the issue of foreign aid effectiveness from a different perspective

that focuses on “targeted aid”. I underscore that analysis of an economic outcome

should be performed with respect to the amount of foreign aid actually devoted to

the concerned projects. With that perspective in mind, this research focuses on

foreign aid targeted towards improving trade capacity of developing countries (Aid-

for-Trade or AfT) and analyzes its impact on the export performance of AfT

recipients.

Two important foundational concerns arise when examining the AfT-export

relation: first, why focus on exports of a developing country? And, second, why aid-

for-trade? A plausible answer to the first concern is the increasingly interdependent

nature of the global economy which offers immense possibilities to developing

countries to tap into the benefits of international trade. Amidst growing trade blocs,

international trade agreements, integrating economies, expansion of global

production and distribution networks by multinational enterprises, unprecedented

international capital and labor mobility, and significant developments in

communication and technology, economists believe that emphasizing international

trade and export led growth are vital for uplifting impoverished lives and the

economic health of developing countries. It is well recognized that trade helps a

country’s economic growth (Awokuse, 2006; Hausmann et al., 2006; Rodrik, 2008).2

2 A detailed explanation can be found in Balassa (1978), Feder (1983) and Abott et al (2009).

4

Exporting increases economic activity, which then helps increase the exporting

country’s employment and income. This further results in increased demand for

products, ultimately increasing both production and exports from that country.

Exporting firms are known for being among the most competitive firms within a

country, which is why an increase in exports promotes and spreads economic

efficiency because it improves the production technology used by these firms.3 This

cycle is expected to create a sizeable ripple effect and work as an engine of growth,

thereby helping alleviate domestic poverty.

The second concern, addressing the importance of AfT in relation to exports,

is the well-accepted fact that aid is essential in promoting exports in developing

countries that historically show poor performance in competing for access to

external markets. The analysis of export performance of developing countries has

drawn even more attention in recent years as donors around the globe have

increased the aid amounts targeted to improve trade capacity. AfT has brought both

aid-advocating communities, such as high-income countries, and trade-promoting

organizations, such as the World Trade Organization (WTO) and the Organization for

Economic Cooperation and Development (OECD), together to work on a

collaborative approach to the overall development of developing countries. While

AfT constituted only 18% of total foreign aid in 2006, this figure increased to 30% in

3 An extensive study on firm level productivity and their export choice can be found on Melitz (2003). Additionally, Helpman, Melitz, and Yeaple (2004) show that firm productivity determines their FDI choices as well.

5

2010, a value equal to 130 billion USD. International organizations like the WTO and

the OECD have pushed the AfT agenda by establishing a special taskforce to promote

allocation of trade-related aid from donor countries.4

It should be noted that AfT is not a separate category of foreign aid, but is

simply a term used to refer to foreign aid targeted to improve trade capacity of

developing countries. It differs from the aid to developing countries from donors for

purposes like emergency relief, health and educational development, poverty

reduction, to name a few. While, these projects directly or indirectly impact

economic development of a country, they do not directly impact exports like AfT

does. According to guidelines from OECD,5 the AfT measure is constructed by

summing the total amount of aid flowing into sectors that directly enhance

economic infrastructures and other services expected to promote exports. This

sectoral data are obtained from the CRS database maintained by OECD under

various headings.6 From these broad headings, AfT reflects the sum of aid that is

categorically spent for: (i) trade policy and regulations, (ii) trade related

infrastructure, and (iii) productive capacity building.

The discussion on the relationship between aid and trade emerged long

before the recent AfT initiatives and produced substantial literature explaining the

motives of foreign aid and examining whether the donors or the recipients benefit

4 Hong Kong ministerial meeting of the WTO requested to set up a special taskforce for this matter. 5 OECD: http://www.oecd.org/document/17/0,3746,en_2649_34665_46582545_1_1_1_1,00.html 6 See Appendix A

6

the most from such relations (for example, Morrissey, 2006; Wagner, 2003; Lloyd,

2000). However, these analyses have been performed with respect to total ODA

because a systematic discussion on AfT began only in 2005. But now that the AfT

data are available, it is time to shift the approach used in aid effectiveness analysis

from general ODA to targeted AfT. To that end, this approach is better equipped to

examine whether developing countries are indeed benefitting from the initiatives

taken by developed countries.

Developing countries typically lack resources required to invest in the

elements that would increase export competitiveness, namely export-promoting

policies, infrastructure (roads, ports), creating a favorable business environment

(banking and financial systems), etc. Ideally, awarded AfT is allocated to improving

these sectors (Cali et al. 2011). AfT plays an important role in the aid-trade-

development linkage, thereby justifying the shift in donor interest towards AfT. It is

therefore a natural question to ask whether the support lent by donor countries is in

fact helping the export sectors of developing countries, thus identifying if AfT is

meeting its economic expectations. In that light, this dissertation presents three

essays that analyze export data of 121 AfT-recipient countries over a period of 16

years [1995-2010] and examine the role of AfT on their export performance.

The first essay examines the effectiveness of AfT by studying its impact on

aggregate level of exports, growth of exports, and change in export-GDP ratio using

the dynamic panel framework and system-GMM techniques. A positive and

7

significant impact, as found in this essay, indicates that aid-for-trade is effective in

stimulating overall exports in aid receiving countries. These findings are in marked

contrast with existing results in the area of general aid effectiveness. However, the

targeted aid is found to exhibit diminishing returns, as with any other financial or

capital resource, thus reinforcing the idea of AfT’s positive but limited role in the

expansion of exports by aid recipients.

Favorable gains in aggregate exports as a result of AfT are laudable provided

that key specific sectors do not suffer adversely in the process. Strictly speaking, AfT

cannot be touted as highly effective if there are losses in the key disaggregated

sectors. Therefore, the second essay concentrates on the export performance of

specific sectors in developing countries. Three major sectors are studied—

agriculture, manufacturing, and service. A major component of this study is the

compilation of sector-wise disaggregated aid-for-trade (SAfT) measures based on aid

data and guidelines from OECD’s Creditor Reporting System (CRS). Since exports in

one sector can be correlated with exports in other sectors within a country, a

seemingly unrelated regression (SUR) framework is used to capture the

interdependence among various sectors in an explicit way, thereby producing

efficient estimates. The results show, in most cases, that sectoral-AfT is effective in

enhancing corresponding sectoral exports, thereby supporting the argument for

allocation of foreign aid to targeted projects.

8

The encouraging results in the previous two cases would not be sufficient to

conclude that AfT is effective if they are achieved at the cost of increased export

variability, because higher export variability is detrimental to overall economic

performance of a country. Hence, the third essay examines the effectiveness of AfT

on reducing the variability of exports for developing countries. The variability is

measured by the sample variance of exports constructed over the period of sixteen

years considered in this dissertation. The data are purely cross-sectional with only

121 available observations, but there are many likely regressors, including AfT, that

impact export performance. This creates an interesting model selection problem,

especially in the absence of any guidance from the existing literature about which

variables to keep and which ones to discard, leading to inaccurate prediction. The

problem is often manifested in the change of statistical significance and even the

signs of the regressors, depending on the combination of macro-regressors used in

the model. This makes it hard for a researcher to pinpoint the set of predictors that

are statistically robust and have strong predictive power.

Appropriate variable selection, therefore, is one of the most common model

selection problems encountered in econometric applications in the empirical macro-

development area. Although several variable selection techniques and criteria have

been used to explain economic performance in general, the use of such techniques

in the AfT-export literature is almost non-existent. For example: Leamer (1985), Sala-

i-Martin (1997), Sala-i-Martin, Doppelhofer, and Miller (2004), Doppelhofer and

9

Weeks (2011), among other, use models based on analyzing sensitivity of a variable

in a pre-built regression analysis. Some other methods such as Forward Selection

and Forward Stagewise Regression start building models by adding variables to the

model based on their correlation to the dependent variable. But, this study utilizes

Least Angle Regression (LARS), an improved version of Forward Selection and

Forward Stagewise Regression methods, to assess the effectiveness of AfT because it

is less greedy and more efficient compared to the other variable selection methods

(Efron et al., 2004). Therefore, by using LARS in assessing AfT effectiveness, we

expect a parsimonious set of explanatory variables for the efficient prediction of

export variability. The findings indicate that GDP-variance, AfT-variance, and

regulatory quality are the most important variables to predict export variability

across countries. When the variances in GDP and AfT show large increments in the

same time period, on an annual basis in our case, the export variance worsens. But,

even when there is higher AfT-variance, aid recipients can reduce export variability

in the presence of good regulatory quality. This last part of our research supports

the existing literature that the recipient country’s internal economic policies can

enhance aid effectiveness.

Overall results indicate that AfT initiatives of the developed countries have

been fruitful in enhancing trade capacity of developing countries by increasing the

level of exports while reducing its variability. These results are derived from three

separate econometric methods that are meant to address a particular issue in each

10

analysis. The dynamic panel estimation technique used in Chapter 2 focuses on

addressing the possible endogeniety issue, the SUR analysis in Chapter 3 captures

possible inter-sectoral correlations, and the LARS used in Chapter 4 addresses a

variable selection problem. All three analyses producing positive results is a strong

evidence in support of such targeted aid, thereby, suggesting a potent channel of

economic expansion and growth. From a development perspective, this study

highlights AfT’s role in improving both the level and variance of exports, thereby

improving international market share and supporting, in the long run, steady growth

in its key economic sectors. However, these positive results are found to exist when

proper economic policies are in place, particularly advantageous regulatory quality.

Therefore, from an aid policy standpoint, it is important for donors to recognize the

value of AfT and allocate aid by targeting specific sectors, and review internal

economic policy environment of aid-recipient countries. Furthermore, because AfT’s

effectiveness is highly dependent on a low AfT-variance, it would be prudent for

donors to be consistent in allocating foreign aid.

11

CHAPTER 2

AID-FOR-TRADE AND AGGREGATE EXPORT PERFORMANCE OF DEVELOPING COUNRIES: A DYNAMIC PANEL ESTIMATION

2.1. Introduction

This paper examines the effectiveness of foreign aid in developing countries

by focusing on aid that is specifically “targeted” to improve trade sector

performance of developing countries, referred to in the literature as Aid-for-Trade

(AfT). The effectiveness of AfT is examined by studying its impact on the level of

total exports by using three different measures of AfT. Additionally, limitations of

AfT are analyzed by studying the diminishing returns of AfT on export levels.

Robustness of these results is checked by using two other measures: export growth

and change in the export-GDP ratio.

Existing literature studies the impact of overall Official Development

Assistance (ODA) on the trade relationship between donor countries and recipient

countries, with special attention to donor exports (Morrissey 2006, Wagner 2003

and Lloyd 2000). Yet in the foreign aid literature it is well recognized that the

majority of such general ODA is spent on public consumption, and is not entirely

funneled to specific projects that help improve export-promoting sectors. Therefore,

contrary to generally accepted viewpoints on measuring the effectiveness of foreign

aid, this research examines the effectiveness of aid specifically in relation to the

recipient’s exports and how aid that is targeted to promote trade sectors can

12

improve the recipient country’s export capacity. The major contributions of this

paper are discovering export patterns in relation to aggregate level of AfT and

determining if there is a positive impact on recipient exports. Specifically, this

includes (i) a focus on export performance of aid recipients, (ii) the consideration of

various AfT categories based on donor type (bilateral vs. multilateral) and end use of

AfT (narrow vs. broad), (iii) an analysis of possible diminishing returns to AfT on

export levels, (iv) a consideration of the additional measures of export growth and

export-GDP ratio beyond the level, (v) the use of more recent and updated AfT

dataset, and (vi) an empirical analysis with the appropriate cutting-edge

econometric technique (system-GMM) to address possible endogeniety problems.

By studying the aggregate level of exports related to AfT, this research

supports that aid is a critical component for a developing country’s growth and

wellbeing at least initially. A country’s ability to promote economic growth is

dependent on its export capacity (Awokuse 2006, Hausmann et al. 2006, Rodrik

2008). As of 2010, total AfT increased to 130 billion USD with the expectations that

the more exports increase the more the impact on economic activity and therefore

the greater the increases in a country’s employment and income. This further results

in increased demand for products, ultimately increasing production and exports

from that country. Exporting firms being among the most competitive firms within a

country, an increase in exports promotes economic efficiency by improving

production technology used by these firms. This cycle is expected to create a

13

significant ripple effect and work as an engine of growth, thereby helping alleviate

domestic poverty. It is also well understood that aid is essential in promoting exports

in developing countries that historically show poor performance in competing for

access to external markets and selling their products.

While the study strongly supports continued growth in exports due to AfT,

there is indication to dampen the commonly held belief that more aid is increasingly

better. In other words, there seems to be evidence to support diminishing returns to

AfT—while more aid has favorable effects, these impacts are positive but

diminishing in magnitude. This broadly defines the nature of AfT effectiveness and

does not mean that aid is unproductive. There is overwhelming support that

continued AfT is needed to most developing countries to uplift their living standards

and further self-sufficiency. Especially, developing countries typically lack capital

required to invest in elements that would increase export competitiveness, namely

export-promoting policies, infrastructure (roads, ports), creating a favorable

business environment (banking and financial systems), etc. Ideally awarded AfT is

allocated to improve these sectors (Cali et al 2008). Thus, AfT plays an important

role in the aid-trade-development linkage, thereby justifying the shift in donor

interest towards AfT. It is therefore a natural question to ask whether the support

lent by donor countries is in fact helping the export sectors of developing countries,

establishing if AfT is meeting its economic expectations.

14

The remainder of the paper is organized as follows: Section 2 gives an

overview of the literature on foreign aid’s effectiveness and details why AfT is

expected to produce favorable export performance in developing countries. Section

3 outlines the methods used to construct AfT and export measures, as well as

discusses the estimation techniques. Section 4 explores the empirical findings.

Section 5 concludes with discussions.

2.2. Conceptual Framework for Analyzing Aid-for-Trade

Donor countries claim to provide foreign aid for a variety of reasons,

including the direct improvement of the social, political and economic situation of

the recipient country. However, aid indirectly improves one or more of these same

areas in the donor country (Alesina & Dollar 2000). Also, the recipient countries

expect general foreign aid to have a significant impact on their overall economic

condition regardless of bilateral or multilateral assistance. Scholars, however, have

not found a significant impact of foreign aid on growth regressions (Moyo 2009;

Rajan 2008; Easterly 2006). Doucouliagos & Paldam (2009) present a more

comprehensive survey of this literature and find out that foreign aid in the last forty

years has not produced favorable results in general. However, there are studies that

show positive impact of foreign aid in the presence of special conditions, such as

15

improved institutional and policy environment (Chong et al 2009; Sachs et al. 2004;

Burnside & Dollar 2000).

Donor countries expect AfT to generate a significantly positive impact on

exports in pinnacle aid-receiving countries, even when sound institutional policies

may be lacking, based on three main arguments. First, developing countries lack

access to international markets. To address this problem, donor countries not only

allocate AfT to help formulate policies and regulations to promote trade in aid-

recipient countries; they also provide some of these developing counties

preferential access to the donor’s market. Second, developing countries face supply-

side infrastructure constraints—roads, ports, etc. — to access other markets. AfT is

clearly designed to help the developing countries improve their export capacity and

reduce these supply-side constraints by targeting aid to transportation,

communication, energy, etc. Up to 60% of total AfT is spent on improving

infrastructure. Lastly, a significant amount of AfT (up to 44%) is spent on business

promotion and banking services.7 This helps increase domestic production, thus

positively impacting exports from these recipient countries.

These conceptual arguments make a clear case that developing countries are

receiving financial assistance to invest in sectors that would otherwise be neglected

in a low-income country due to its weak tax base, as well as the fact that a majority

7 See Appendix A for breakdown of AfT categories.

16

of its collected revenue is spent primarily on consumption sectors (Arellano et al,

2009). To assess the AfT-export effects three testable hypotheses can be articulated:

(i) AfT increases the level of exports from aid-recipient countries, (ii) AfT impacts the

growth of exports positively over time, and (iii) an increase in the level and growth

of exports should result in an increase in the size of total exports compared to the

size of the economy. These represent different strengths of the effects of targeted

aid on exports, in terms of export level, its growth, and change relative to GDP.

Having all three positive and significant would depict a highly optimistic picture of

the effectiveness of AfT. These in addition to an examination of diminishing returns

offer a more detailed and meaningful view of the role of AfT in export promotion.

The above expectations sound overly ambitious at a time when effectiveness

of foreign aid on economic growth is generally questionable. The assumption here is

that AfT does not have to show a positive impact on economic growth directly for it

to be considered effective because it indirectly leads to long-term economic

prosperity by positively influencing exports. The key here being exports rather than

economic growth, the idea being that enabling export capacity may eventually

establish an engine of growth that leads to higher GDP. Indeed, the limited existing

research on AfT shows rather mixed results. The nominal amount of research is due

to the fact that rigorous discussion on the AfT agenda started only during the WTO’s

Hong Kong ministerial meeting in 2005. This meeting established a special task force

17

to promote AfT, even though there was demand for increased trade-related

assistance in the WTO’s Doha meeting in 2001 (OECD, 2006).

Before the systematic discussion of AfT, previous academic literature

concentrated on examining the relationship between foreign aid and trade with

respect to general foreign aid (ODA). The data present mixed results; Wagner (2003)

shows that donors’ exports increased more than recipient’s exports as a result of

providing increased foreign aid (ODA) to developing countries. On the other hand,

Lloyd et al (2000), using ODA as the measure of foreign aid, found that recipient

countries’ exports increased, while increases in donor exports were an exception,

rather than the norm. More recent literature, such as Winters (2010), emphasizes

the need to examine the effectiveness of foreign aid based on its targeted sector.

Similarly, Johansson and Petersons (2011) study a bilateral trade relationship

between the donors and recipients with respect to AfT. They find that an increase in

flow of AfT increases exports of donor as well as recipient countries.

The present study differs from the previous literature examining the AfT-

Trade relationship in several ways. First, most previous studies have examined the

impact of AfT on bilateral exports. They found that exports from aid recipients

increase to aid donors, but only selectively (Brazys, 2010). I argue that it does not

matter where the exports are going; all that matters is whether or not the export

capacity of the recipient is improving as a result of the aid received. Second, this

18

study uses a more recent and updated dataset, giving a better picture of AfT’s

influence on exports from recipient countries. Third, this study disaggregates AfT

measures into several categories largely unaddressed by previous studies revealing

information that would other be masked in an aggregate aid-trade analysis.

Although previous papers show the relationship between the level of AfT and

the level of exports, they do not explain the limitations of such aid and/or the

channels through which the gains in exports were achieved. This paper fills that gap

by examining the diminishing returns to AfT and analyzing the relationship between

AfT and two aspects of export performance, in addition to the level of exports, in an

effort to better explain the economic effectiveness of foreign aid. Thus, this paper

examines whether AfT helps export growth. It also explores patterns between

exports and GDP by analyzing annual shifts in the export-GDP ratio due to AfT.

2.3. Data and Empirical Analysis

This study analyzes an annual level longitudinal dataset of 121 AfT-recipient

countries over a period of 16 years [1995-2010]. Although discussion on AfT began

with a design to help the least developed countries, there are a number of middle

income countries benefitting as well.8 As such, this study considers countries

categorized as low-income and middle-income by the World Bank. The export data

8 See Appendix C for list of countries included in the study.

19

are obtained from United Nation’s Conference on Trade and Development

(UNCTAD) except for the service exports, which are obtained from World

Development Indicator database maintained by the World Bank. The AfT data are

obtained from OECD’s Creditor Reporting System (CRS). Other control variables

included in the analysis are GDP per capita, money supply, exchange rate, trade

openness, and institutional variables like control of corruption and regulatory

quality. Trade openness is measured by trade freedom and the data are obtained

from Heritage Foundation; GDP per capita, broad money supply and exchange rate

are obtained from World Development Indicator database, and finally the

institutional variables are produced by Kaufmann, Kraay, and Mastruzzi (2010) and

are obtained from World Governance Indicator database maintained by the World

Bank.

2.3.1. Measuring Aid-for-Trade

The main variable of interest in this analysis is Aid-for-Trade, denoted by AfT

in the regression model. It should be noted that AfT is not a separate category of

foreign aid. The donor countries and multilateral agencies provide aid to developing

countries for purposes like emergency relief, health and educational development,

poverty reduction, to name a few. These projects directly or indirectly impact

economic development of a country. Following the guidelines from the Organization

20

for Economic Cooperation and Development (OECD)9, the AfT measure is

constructed by summing the amount of aid flowing into sectors that directly

enhance economic infrastructures and other services expected to promote exports.

These sectoral data are obtained from the CRS database maintained by OECD under

different headings.10 From these broader headings, total AfT reflects the sum of aid

that is categorically spent for: (i) trade policy and regulations, (ii) trade related

infrastructure, and (iii) productive capacity building.

In this paper, the impact of AfT on exports is analyzed with respect to the

broader measure, total AfT, as well as some narrower measures. The narrower

measures are classified (a) in terms of their end use and also (b) in terms of the

nature of donors. Based on the end use of AfT, the first narrower measure (AfT1)

reflects the amount of aid that is spent solely on trade policy and regulations. This

consists of items ranging from administrative management, trade education and

training, trade facilitation, and regional and multilateral trade negotiations and

agreements. A slightly broader measure, AfT2, consists of AfT1 plus the aid that is

provided for trade-related infrastructures – such as transport and storage,

communications and energy. Finally, the broadest measure, total AfT, consists of

AfT2 plus aid provided for productive capacity building which includes sectors

ranging from agriculture to business and banking services, as shown in Appendix A.

9 OECD: http://www.oecd.org/document/17/0,3746,en_2649_34665_46582545_1_1_1_1,00.html 10 See Appendix A

21

All of these categories are not necessarily expected to have the same impact since

that would depend on the export sector one is analyzing.

This paper also distinguishes the impact of bilateral versus multilateral AfT.

Based on the origin of AfT, it is classified as bilateral AfT if the aid given by the

government of a donor country directly to the government of a recipient country.

On the other hand, if AfT reaches the recipient through an agency like World Bank,

WTO or any other UN affiliate, it is included in the multilateral category. This

distinction is important because the impact of aid is shown to be different

depending on its origin due to economic motives (for example, being able to export

more to the recipients) as well as the level of implementation efforts and

monitoring; hence, the effectiveness of aid in terms of promoting recipient

countries’ exports may differ depending on the source of that aid (Svensson, 2000).

2.3.2. Measuring Export Performance

Based on conceptual arguments and testable hypotheses explained in

section 2, this paper analyzes export performance of developing countries from

various perspectives. First, the paper examines the impact of AfT on the export

levels. Export levels are measured by the dollar value (adjusted for inflation) of total

exports: goods and services added together. Next, the paper examines if there is any

evidence of diminishing returns to AfT, a concept well established in economics and

22

well documented in empirical studies pertaining to economic growth. Examination

of such concept in this type of specific analysis would be interesting. To that end, the

paper analyzes the impact of the squared term of AfT on export. Also the size of AfT

in relation to the size of the economy, as measured by the ratio of AfT received by a

country to its GDP, is used. While analysis of export levels is relevant, this paper

further examines if AfT has any effect on export growth. These measures combined

provide a robust analysis of the relationship between AfT and exports. Accordingly,

these export performance measures are calculated using equations (2.1) and (2.2).

𝐸𝑥𝑝𝑜𝑟𝑡 𝐺𝑟𝑜𝑤𝑡ℎ = ln(𝑋𝑖𝑡) − ln�𝑋𝑖,𝑡−1� (2.1)

𝑆ℎ𝑎𝑟𝑒 𝐶ℎ𝑎𝑛𝑔𝑒 = 𝑋𝑖𝑡𝐺𝐷𝑃𝑖𝑡

− 𝑋𝑖,𝑡−1𝐺𝐷𝑃𝑖,𝑡−1

(2.2)

Where, X denotes the export volume measured in US Dollar, ‘i’ represent 121

AfT-receiving developing countries and ‘t’ represent yearly observations from 1995

through 2010.

2.3.3. Model Specification and Estimation Technique

Since current exports are likely to depend on past exports, the paper uses a

dynamic panel specification while analyzing the impact of AfT on the level of

23

exports. The specification also includes important control variables that are part of

most trade regressions. The empirical model takes the following form:

𝑋𝑖𝑡𝑘 = α𝑖 + β1𝑋𝑖,𝑡−1𝑘 + β2𝐴𝑓𝑇𝑖𝑡 + β3𝑌𝑖𝑡 + β4𝑇𝑂𝑖𝑡 + β5𝐹𝐷𝑖𝑡 + β6𝑋𝑅𝑖𝑡 + β7𝐶𝐶𝑖𝑡 + β8𝑅𝑄𝑖𝑡 + ε𝑖𝑡

(2.3)

Where, 𝑋𝑖𝑡𝑘 represents the log of export level of sector k from an aid recipient

country ‘i’ to the rest of the world at time ‘t’. 𝐴𝑓𝑇𝑖𝑡 denotes the log of Aid-for-

Trade, the main variable of interest impacting exports. 𝑌𝑖𝑡 is the log of real GDP per

capita of the exporting country. 𝐹𝐷𝑖𝑡 represents a measure of financial development

using real broad money supply as a proxy, which plays a crucial role in financing

export promoting businesses. 𝑋𝑅𝑖𝑡 represents the real exchange rate of domestic

currency vis-à-vis US Dollar. 𝐶𝐶𝑖𝑡 and 𝑅𝑄𝑖𝑡 represent control of corruption and

regulatory quality, respectively. These institutional measures are relevant in this

type of study as their quality determines the effectiveness of foreign aid. Finally,

𝑇𝑂𝑖𝑡 represents the trade openness measured by trade freedom of that country.

This is an index measuring import tariff and quota, voluntary export restraint, etc.

So, it is expected to proxy for both import and export openness. All of these

variables are expected to impact exports positively.

Given the nature of foreign aid flows, there exists possible reverse causality

between exports and AfT, leading to the problem of endogeniety. The presence of

lagged dependent variable in a panel setup can also produce endogeneity (as is well

24

known). To address these issues, System Generalized Method of Moments (GMM) is

used for estimation (See Arellano and Bover 1995). This technique uses instrumental

variables from the system that consist of the exogenous and predetermined

regressors in the regression model. This method also provides more efficient

estimators compared to Arellano-Bond (1991) GMM estimation.11

Similarly, export growth and change in export-GDP ratio are also analyzed

using System-GMM. These two regressions are analyzed using the explanatory

variables in level as well as in growth terms for AfT, exchange rate and money supply

to have a one-to-one matching of the variables. These new variables are constructed

following the same method used in the construction of export growth and the share

as in (1) and (2).

2.4. Empirical Results

This section explains the results obtained from empirical estimations. The

first sub-section examines the impact of five different categories of AfT on total

exports, followed by diminishing returns to Aft; and the third sub-section examines

the impact of total AfT on yearly growth of total exports, along with annual changes

in the share of exports to GDP.

11 Econometric estimation is performed using STATA-11

25

2.4.1. Explaining Export Levels

The first column in Table (1) reports the impact on export levels due to AfT1,

the narrowest measure of AfT in terms of end use. Similarly, the second column

reports results due to AfT2, and the third column presents the total AfT as the aid

variable. All of these measures have positive and highly significant coefficients on

the level of total exports, except in the case of AfT1. Finally, Columns (4) and (5)

analyze the differences in impact with respect to classification of AfT based on their

origin, bilateral AfT and multilateral AfT. Both the bilateral and multilateral AfT have

a similar impact on aid-recipient total export levels: a positive and statistically

significant impact. One striking observation is the favorable impact of total Aid-for-

Trade (Column 3) with high statistical significance. This demonstrates that

developing countries need comprehensive assistance to increase their level of

exports to the global market.

One possible explanation for the adverse impact reflected in the AfT1

measure could be that the removal of trade barriers may overexpose domestic

exporting firms to competition from newly entered importing firms, impairing

exporting firms’ overall production potential and thus and export capacity. In fact,

these results are consistent with the negative sign obtained with respect to trade

openness. At first glance, these results seem to be counterintuitive. Nevertheless,

increases in imports can actually hurt infant industries in an aid-recipient country.

26

Opening up markets to foreign competition in those developing countries means

that start-up firms possessing poorer technologies and capital fare adversely when

competing with international firms better equipped in technology, human resources

and physical capital.

This paper is not, by any means, arguing that developing countries should not

employ trade-openness. However, there must be proper financing and facilitation

options available to exporters in aid-recipient countries so that they may become

more competitive within their chosen markets. The results merely indicate that

removing trade barriers alone is not sufficient for these developing countries; other

complementary programs (infrastructure, productive capacity), are equally

important to boost their level of exports. This is a long-term process, and there must

be other policy tools to go hand-in-hand with trade barrier removal in order to reap

the benefits of trade openness when embraced by aid-recipient countries. This

argument is supported by the positive and significant coefficients obtained with the

addition of aid for infrastructure (category 2 of AfT) and productive capacity building

(category 3 of AfT). Improvement of economic infrastructure is a major objective of

AfT, including construction and improvement of roads, ports, telecommunication

and energy networks, because such infrastructure improvements have proven to

promote the overall level of exports.

27

Other control variables in most situations produce expected signs. Income of

exporting countries exhibits a positive and significant impact on exports in all

regressions. This supports the fact that higher-income developing countries export

more than lower-income countries, possibly due to the higher level of production,

technology, and human capital in those countries. Similarly, an increase in the

exchange rate (i.e. - devaluation of domestic currency vis-à-vis USD) is found to

improve exports of total goods and services. This is consistent with the general

theoretical assumptions of currency devaluation.

The impact of financial development, measured by the level of money supply

on the level of total exports, does not show a robust result. Beck (2002) argues that

financial development has a positive impact on exports. They analyze the finance-

trade relationship using credit to private sectors by financial intermediaries as the

measure of financial development. Broad money supply is the best available

measure of financial development for the sample of countries considered in this

analysis. So, the positive coefficients are consistent with the theory and previous

empirical papers. However, a possible explanation for the negative impacts

observed, in the analysis of our sample of developing countries, may be that an

increase in money supply may not translate to better access to credit and financial

instruments that are likely to be conducive to export promotion; more money supply

may simply mean better exchange opportunities rather than better production and

export prospects in those countries.

28

Next, the impact of control over corruption as a measure of institutional

quality displays mixed results. While reduction in corruption shows a negative

impact when analyzed with disaggregated AfT, the results are insignificant when the

total amount of AfT is considered. The negative sign may mean that employees at

ports and licensing offices used to clearing goods and licenses for a fee are causing

delays because they now lack the incentive to expedite their services. This should

prove to be a short-term consequence of disrupting corruption; in the long run, one

would expect that control of corruption helps improve exports, along with other

economic sectors.

Finally, as expected regulatory quality measuring the ability of the

government to create business-friendly policies and regulations has, for the most

part, a positive and significant impact.

2.4.2. Explaining Diminishing Returns to Aid-for-Trade

Tables (2) and (3) show results pertaining to diminishing returns to AfT. First,

the diminishing returns are analyzed with respect to squared term of AfT. Taking

into account the increase in overall AfT from 18% of ODA in 2006 to 29% in 2010

(see Appendix A), a negative coefficient on AfT-squared implies that AfT exhibits

diminishing returns, which would suggest that aid recipients in the initial phase

experience a more substantial increase in export volume per dollar of aid than

29

subsequently with greater levels of aid. Diminishing returns, as evidenced here, thus

puts an upper bound on the effectiveness of AfT. This indicates that while the

targeted aid is effective in enhancing export capacity, AfT on its own can only have a

limited impact, with an elasticity of less than unity. That is, a one percent increase in

AfT will have a less than one percent increase in exports. From a policy perspective,

this means that AfT alone may not be relied upon as the only tool required for

developing countries; other resources are required for export promotion and

expansion simultaneously. These programs might range from human capital

improvement to upgrading of manufacturing and production technologies. This has

an important implication to the policy makers involved in crafting AfT related

projects in both the aid donating as well as aid receiving countries.

Similarly, a negative coefficient observed on the AfT-GDP ratio shows that

developing countries perform poorly when they are heavily dependent on foreign

assistance. The findings are consistent with the aid-dependency theory in general.

One of the popular explanations for such a case is the weakening of political

institutions (governments in most cases) because of less accountability they have

towards their citizens. In that case, the additional foreign assistance targeted for

trade related activities might actually be diverted to other consumption sectors

because the governments have less incentives to invest that aid in improving trade

favoring policies, trade related infrastructures, and harnessing business friendly

environment, which are the pillars for developing a competitive trade sector.

30

Combined these two findings, while we do not refute the positive gains

obtained from AfT, the results contradict the generally held perception that

increasingly more aid is always better. Hence, these predictions pose a challenge to

the aid donors to strategically identify the actual amount of foreign aid needed

based on two criteria: first, the resources available to a developing country and

second, the size of its economy. A deeper analysis to identify an appropriate timing

and amount of AfT that would produce optimum results would be an interesting

future line of research.

2.4.3. Explaining Export Growth and Change in Export-GDP Ratio

From the first set of regressions, it is clear that level of AfT favorably impacts

the level of exports. Yet curiosity looms over the dynamics of exports over time.

With the onset of AfT in a given country, an increase in export levels would be worth

more if accompanied by positive growth of exports. In order to measure such an

export dynamic, two measures of export are analyzed: the growth of export and the

change in the export-GDP ratio. The first column in Table (4) reports the coefficients

with respect to explanatory variables in levels. Similarly, the second column reports

the coefficients with respect to explanatory variables in growth. Both of these

specifications show that AfT has a positive and significant impact on growth of

exports, controlling for all other variables influencing exports. Again, higher-income

31

countries gain stronger export growth compared to their counterparts. Depreciation

of an exporter’s currency vis-à-vis USD helps its export growth, while a growth in the

exchange rate does not have any significant impact on exports. Money supply has a

negative impact, while growth of money supply has positive impact. Openness,

again, has a negative impact on export growth.

A change in the share of export to GDP ratio from one year to the next is

used as an alternative measure to study export dynamics. The share of exports in

GDP maintains a special meaning because an increase in exports relative to GDP

means a more favorable export environment in a given country. Table (5) reports

the coefficients for variables on both the level and the growth. Both specifications

show that AfT has a positive and significant impact in increasing the difference in the

export-GDP ratio. These results are consistent with the positive impact of AfT on

export growth. Once again, income strongly influences the export-GDP ratio’s

growth from one period to the next. The remaining explanatory variables are

consistent with the results in export growth. The institutional variables do not show

robust impact since their significance depends on the specific model in question.

2.5. Conclusions with Discussions

This paper contributes to the literature by scrutinizing the effectiveness of

foreign aid on export performance of aid recipient countries in a comprehensive

32

way. The impact of Aid-for-Trade on export performance is analyzed at the

aggregate level of exports considering various categories of AfT based on their origin

as well as end use. Although the narrowest measure of AfT shows negative impact

on the level of exports, overall results show that total exports from aid-recipient

countries significantly increase as a result of AfT. Thus, the analysis performed in this

research with various categories of AfT (narrower vs. broader) unfolds the aid-trade

relation in more detail and more thoroughly which would otherwise have remained

masked. By using disaggregated measures for AfT, no new assumptions or

conceptual arguments were made in examining foreign aid; this makes our findings

of greater importance as donors provide additional AfT for the greatest impact in

the future.

This paper also examines the impact of aid on the annual export growth

pattern, as well as change in the share of export to GDP. Both of these measures

exhibit a favorable impact as a result of increased AfT. Overall, this paper finds that

AfT has a positive impact on export performance, which further calls for increased

export-targeted foreign-aid to developing countries since many countries still have

yet to realize widespread benefits that could be achieved with improved

employment and incomes that come with a competitive trade sector.

These results are based on the average impact of aid on all AfT recipients.

Case studies of individual countries would likely provide a clearer picture of the

33

realities facing these countries, including the limits imposed by diminishing returns.

Future work must be geared toward such an approach. Improved export

performance of developing countries depends on many more factors than the

financing of projects by foreign aid; effectiveness of foreign aid will be noteworthy

when proper economic environment is in place in developing countries to promote

exports in the long run.

34

CHAPTER 3

SECTORAL AID-FOR-TRADE AND SECTOEAL EXPORTS: A SEEMINGLY UNRELATED REGRESSION ANALYSIS12

3.1. Introduction

This chapter examines the effectiveness of bilateral and multilateral foreign

aid disbursed for the purpose of export promotion in developing countries (aid-for-

trade or AfT). We focus on three major sectors, namely manufacturing, agriculture,

and services, to investigate the effects of various AFT measures (sector-specific AfT

or SAfT) on the levels of exports in the above three sectors. Because sectoral exports

can very well be correlated to each other, a Seemingly Unrelated Regression (SUR)

model proposed by Zellner (1962) is used to produce more efficient estimates. The

three major contributions of this essay are: (i) construction of Sectoral-AfT based on

recommendations of the Organization for Economic Cooperation and Development’s

(OECD) Creditor Reporting System, (ii) focus on the impact of sectoral AfT on

sectoral exports, and (iii) the use of the SUR method to capture inter-sectoral

correlations.

Systematic discussion of AfT began in 2005 with the formation of a taskforce

by the World Trade Organization (WTO) to pinpoint projects that would enhance

export performance of developing countries. The committee submitted a report that

outlined thirteen major categories of foreign aid that directly impact exports from

12 Research based on this chapter has been submitted for publication in Economics Bulletin.

35

developing countries. These aid categories range from improvement of trade

policies to infrastructure development.13 One of the major recommendations of the

taskforce was to identify comparative advantage at the country level as well as

sector level and allocate AfT accordingly to respective sectors (WTO, 2006). Since

then, various institutional reports have attempted to analyze the effectiveness of

AfT; for example, WTO and OECD jointly publish a biennial progress report on the

AfT initiative (OECD/WTO (2011)).14 However, systematic empirical research (at the

cross-country level) focusing on the effectiveness of AFT on export promotion of

developing countries is surprisingly limited. Morrissey (2006), Wagner (2003) and

Lloyd (2000) focus on the effects of bilateral aggregate foreign-aid on bilateral

exports, while Cali et al (2011), in a case study, suggest how small and vulnerable

Caribbean economies would benefit from AfT by reducing the cost of trading.

Johansson and Peterson (2011), using a broader sample of countries, examine the

impacts of total bilateral AFT on total bilateral exports of developing countries. The

present study is the first attempt to examine the impacts of sector-wise AfT on

sector-specific exports of developing countries and capture sectoral

interdependence by adopting a SUR analysis.

13 See Appendix B for a complete list of aid categories included in Aid-for-Trade. 14 The reports mainly focus on descriptive statistical analysis, examples, and country-specific case studies.

36

3.2. Data and Measuring Sectoral Aid-for-Trade

This study analyzes annual export data of 121 AfT-recipient developing

countries over a period of 16 years [1995-2010]. Although discussion on AfT began

with a design to help the least developed countries, there are a number of middle-

income countries benefitting as well.15 As such, this study considers countries

categorized as low-income and middle-income by the World Bank. The export data

are obtained from United Nation’s Conference on Trade and Development (2012),

except for the service exports which are obtained from the World Development

Indicators (2011) maintained by the World Bank. The aid data are obtained from

OECD’s Creditor Reporting System (2012). Other control variables included in the

analysis are per-capita value-added in corresponding sectors, financial development

(using bank private credit to GDP ratio as a proxy), exchange rate, trade openness

(using trade freedom as a proxy), and institutional variables, such as control of

corruption and regulatory quality. Per-capita value-added and exchange rate are

obtained from World Development Indicators (2011), measure of financial

development is obtained from Global Financial Development (2011), and the

institutional variables – control of corruption and regulatory quality are obtained

from Worldwide Governance Indicators (2011) – all three databases being

maintained by the World Bank. Finally, trade openness is obtained from Heritage

Foundation (2011).

15 See Appendix A for list of countries included in the study.

37

A major contribution of this study is the construction of Sectoral Aid-for-

Trade (SAfT) measures, i.e. aid allocated for export promotion and development of

agricultural, manufacturing, and service sectors, exclusively. In the absence of any

clear guidelines regarding their construction, this study proposes three possible

measures of SAfT for each sector based on their potential scope; the narrowest

measure includes the category of aid that primarily influences the respective sector

of interest, a slightly broader measure includes aid categories that possibly influence

more than one sector, whereas the broadest measure includes aid categories that

are important for exports in general. The narrowest SAfT for the agricultural sector

includes aid allocated for sub-categories such as agriculture, forestry, and fishery.

Similarly, the narrowest measures for aid for the manufacturing sector includes

industry, construction, transportation, communication, and energy; and the aid for

service sector includes banking and business services, communication, energy, and

tourism. The two broader measures of aid for each sector include additional aid

categories ranging from trade policies to other infrastructure development (See

appendix B for details on each measure).

3.3. Model Specification and Estimation Technique

For any given country, exports in a sector can well be correlated with exports

in other sectors and these sectoral exports are also likely to be affected by common

macroeconomic shocks. The study therefore, uses a Seemingly Unrelated

38

Regression (SUR) model as proposed by Zellner (1962) while analyzing the impact of

SAfT on the levels of sectoral exports. In our seemingly unrelated system of

equations, we have three equations—each for one sector (agriculture,

manufacturing, and service), and sector specific exports appear as the response

variable for the corresponding equations. Also, for each equation, we have

corresponding SAfT and sector-specific value-added as the main regressors.

Additionally, we include some common macroeconomic variables in each equation.

As such, sectoral aid and value added by each sector differ in the system of SUR

specification, while other control variables remain unchanged throughout the

system of equations. The Breusch-Pagan (1980) Test of Independence rejects the

null hypothesis of zero correlations across sectoral equations, with p-values around

0.000. Hence, Zellner’s (1962) SUR estimation is an appropriate method to apply for

the present study enabling us to obtain efficient estimates. Two sets of analyses are

performed. In the first set, we study the impacts of average SAfT on average sectoral

exports (averaged over the 16-year period). Equation (3.1) below captures such

purely cross-sectional relations.

𝑋𝑖𝑘 = α + β1𝐴𝑓𝑇𝑖𝑘 + β2𝑉𝐴𝑖𝑘 + β3𝑇𝑂𝑖 + β4𝐹𝐷𝑖 + β5𝑋𝑅𝑖 + β6𝐶𝐶𝑖 + β7𝑅𝑄𝑖 + ε𝑖𝑘 (3.1)

Where, k=1, 2, 3 represent the three sectors in our study. The second set of analysis

is performed by using longitudinal data for 121 countries over 16 years, regressing

levels of sectoral exports on sectoral AFT. The corresponding empirical model, with

α𝑖 as a random effect, appears in the following equation.

39

𝑋𝑖𝑡𝑘 = 𝜆 + β1𝐴𝑓𝑇𝑖𝑡𝑘 + β2𝑉𝐴𝑖𝑡𝑘 + β3𝑇𝑂𝑖𝑡 + β4𝐹𝐷𝑖𝑡 + β5𝑋𝑅𝑖𝑡 + β6𝐶𝐶𝑖𝑡 + β7𝑅𝑄𝑖𝑡 + α𝑖 +

ε𝑖𝑡𝑘 (3.2)

Where, 𝑋𝑖𝑡𝑘 represents the log of export of sector k from an aid recipient country ‘i’

to the rest of the world at time ‘t’. 𝐴𝑓𝑇𝑖𝑡 denotes the log of Sectoral Aid-for-Trade,

the main variable of interest impacting exports. Another important explanatory

variable in the analysis is the sector specific value added – 𝑉𝐴𝑖𝑡 – measured as the

log of real per capita value-added by each sector of the exporting country. 𝑇𝑂𝑖𝑡

represents the trade openness measured by trade freedom of that country. This is

an index measuring import tariffs and quotas, voluntary export restraints, etc. So, it

is expected to proxy for both import and export openness. 𝐹𝐷𝑖𝑡 represents a proxy

of financial development measured by the ratio of bank private-credit to GDP, which

plays a crucial role in financing export promoting businesses. 𝑋𝑅𝑖𝑡 represents the

real exchange rate of domestic currency vis-à-vis US Dollar. Finally, 𝐶𝐶𝑖𝑡 and 𝑅𝑄𝑖𝑡

represent control of corruption and regulatory quality, respectively. Such

institutional measures are important in this type of study as their quality determines

the effectiveness of foreign aid. All of these variables are expected to impact exports

positively.

3.4. Empirical Results

The results obtained from regressing purely cross sectional observations as

specified in equation (3.1) are presented in Table 6. The first half of the table

40

presents the estimates based on the narrowest measure of SAfT for each sector; the

second half of the table presents the results based on the broadest measure for

each sector. The Sectoral-AfTs for all three sectors have positive and highly

statistically significant coefficients in all cases, as expected. The empirical analysis

was also performed using the middle measure, broader SAfT, but is not reported

here for brevity; however the results do not differ from those obtained using the

broadest measure of SAfT.

Similarly, Table 7 summarizes the results obtained by regressing the data in a

panel setup as specified in equation (3.2). Although the agricultural sector shows a

negative impact with respect to narrowest measure of agricultural SAfT, all other

measures seem to impact respective sectoral exports positively and statistically

significantly. Similar analysis was performed using the broader measure of SAfT,

without changing the results compared to the broadest measure.

Sectoral value added seems to be an important explanatory variable in the

exports regression, as shown by the associated positive and highly significant

coefficients throughout all the specifications. Other control variables show mixed

impacts on sectoral exports depending on the model used. Trade openness does not

show any strong statistical significance on the sectoral exports in cross-sectional

study but shows negative impact on agricultural and manufacturing exports, and a

positive impact on service exports. Similarly, the devaluation of domestic currency

shows no statistical significance when the data are averaged over time (Table 6), but

41

have mixed results in a panel setting (Table 7), and the results depend on how SAfT

is measured. An improvement in financial sector as measured by bank private-credit

to GDP ratio, mostly shows a positive impact on exports of all three sectors, except

for the agricultural exports under cross-sectional setup. While control of corruption

mostly has negative impact on sectoral exports, it shows a positive impact when the

data are specified as panel and the narrow measure of AfT is used. Finally,

regulatory quality shows a positive impact on all three sectors for a cross-sectional

observation, but the signs flip when the data are setup as a panel.

Overall results show that the estimates on the control variables appear to

change when the observations are reported in a yearly fashion but such differences

disappear when the data are averaged over time. Although, a detailed analysis of

these control variables is beyond the scope of this study, we find that mostly

Sectoral-AfT and sectoral value added have consistently positive and significant

coefficients across equations and across sectors. The results in this study, therefore,

call for a more appropriate allocation of sector-specific foreign aid to developing

countries based on their corresponding needs.

3.5. Conclusions

A major contribution of this study is using the Seemingly Unrelated

Regression (SUR) to analyze the impact of sectoral aid on corresponding sector-

specific exports; this approach produces efficient estimates after taking care of inter-

42

sector interdependence. The study constructs three measures of sectoral aid, which

are then seen to exert a positive and statistically significant influence on sectoral

exports in the key areas of manufacturing, agriculture and service. The results

presented in the study are robust to different model specification and various

measures of sectoral-aid and suggest a favorable role for aid that is specifically

targeted for trade.

Our study also suggests the need for more research in this particular area to

highlight the connections between sectoral aid and exports in the different stages of

economic development. To the extent that trade happens to be a prominent engine

of growth and development for many recently emerged countries, it would help to

identify when the potential effect of aid on trade is sizeable. For example, it would

be useful to know if aid-for-trade is more effective in low-income or middle-income

countries, so that aid resources could be appropriately employed.

43

CHAPTER 4

EFFECTIVENESS OF AID-FOR-TRADE ON REDUCING EXPORT VARIABILITY: SEARCHING FOR BEST PREDICTORS USING LARS

4.1. Introduction

This chapter examines the effectiveness of foreign aid in developing

countries by addressing the variable selection problem encountered in applied

development studies of macroeconomic orientation. In particular, it analyzes the

effect on variance of exports due to foreign aid that is specifically targeted to

improve the export performance of developing countries, referred to in the

literature as Aid-for-Trade (AfT). To that end, the study uses the Least Angle

Regression (LARS) method to suggest appropriate explanatory variables that are

proper fit for the model.

This focus on the analysis of export variability of the aid recipients

contributes to existing literature in two aspects: first, it identifies the form (i.e. level

or variance) of independent variables to be used in the model when the dependent

variable is sample variance of exports; and second, it analyzes whether AfT, in one

form or the other, shows a significant impact in reducing export variability of AfT

recipients. Although the development community has focused on the importance of

AfT to improve the level of exports from developing countries, the role of lower

variance in boosting exports from the same set of countries has not gained much

attention. While past studies, including the results obtained in the previous chapters

44

of this dissertation, show an increase in the level of exports due to an increase in

AfT, the focus here is the impact of AfT on the variability of exports. To be specific,

aid is presumably more effective when it raises the level of exports and does so with

less variability. Higher variance of exports would likely mean reduced well-being

particularly in low export years, despite having a higher average export level as

shown in chapters two and three. This would translate to lower employment and

national income in the adversely affected years. From a conceptual point of view,

given a higher export level, it is obvious that a lower export variance would be

desirable. From a policy standpoint, it makes good sense to seek out targeted aid

that satisfies both criteria — enhances the level of exports, but also keeps export

variability low. From a development perspective, variability of exports would impose

uneven growth on the key sectors and create disruptions that may undermine the

long-run growth process, possibly even leading to loss of international market share

and even causing social unrest within the country. Given weak infrastructure and

new participation in export markets by such aid receiving countries, lower variability

of exports can add to stability and balanced growth.

Empirical analysis of export variability is a somewhat tricky issue, especially

given sparse literature indicating a specific methodology to use when variability is

the dependent variable. This essay incorporates an efficient methodology, LARS, to

show that AfT and export variance are linked in a favorable way. In fact, a challenge

faced by empirical economists studying foreign aid effectiveness is the selection of a

45

proper set of variables that yield an appropriate econometric model with maximum

and consistent predictive power. This challenge is also evidenced in various studies

analyzing the impact of foreign aid on economic growth of the aid recipient

countries. The debate on the effectiveness of foreign aid rekindled when Burnside

and Dollar (2000) reported that foreign aid was more effective in aid recipients with

good monetary, fiscal, and trade policies. However, Hansen and Tarp (2001) show

that foreign aid effectiveness is in fact sensitive to the choice of estimators and the

set of explanatory variables used in the model. Similarly, Easterly et al. (2004) show,

again as a response to Burnside and Dollar, that the robustness of foreign aid on

growth regressions is questionable when the sample of countries and the period of

analysis are altered. Although it is difficult to make a similar comparison across the

literature in the case of robustness of AfT, we can make use of econometric methods

that assess the robustness of AfT across different models. In that light, this is the

first study to my knowledge that utilizes the LARS method in foreign aid literature to

identify whether AfT is one of the appropriate determinants of export variability

from developing countries.

The remainder of the chapter is organized as follows: section 2 outlines an

overview of the literature on variable selection methods; section 3 explains the data

and LARS as the variable selection technique; section 4 explores the empirical

results; and, section 5 summarizes the findings.

46

4.2. Review of Literature

The systematic discussion on AfT began in 2005 with the formation of a

taskforce by the Organization for Economic Cooperation and Development OECD to

recommend the types of projects that would be considered as recipients of AfT. The

committee submitted a report outlining thirteen major categories of foreign aid that

directly impact exports from developing countries. These aid categories range from

improvement of trade policies to infrastructure development. Even though these

recommendations aimed to increase the level of exports and to reduce the volatility

of exports from developing countries, existing studies on AfT effectiveness have also

emphasized the impact of AfT on the level of exports. A case study presented by Cali

et al. (2011) suggested how small and vulnerable Caribbean economies would

benefit from AfT by reducing the cost of trading and, hence, increasing export levels.

Similarly, Johansson and Peterson (2011), using a broader sample of countries,

examined the impacts of total bilateral AfT on total bilateral exports of developing

countries. Both studies found that AfT influences export levels of both donors and

recipients positively.

There are a limited number of recent studies that have shifted the discussion

from analyzing export levels to the examination of export variability, but all use an

indirect approach. Munemo (2011) analyzes the impact of overall foreign aid on

export diversification. Similarly, Haddad (2011) analyzes the role of international

organizations such as the WTO in diversifying exports from developing countries.

47

These articles argue that an increase in diversification of exporting sectors reduces

the volatility of exports; competence in exporting multiple goods and services allows

a country to maintain a balanced export from one period to the next even if one of

the sectors is adversely impacted. However, no research exists that examines the

relationship between foreign aid and export variability directly. So, this study

attempts to fill this gap, measuring the variability of exports by the sample variance.

Mansfield and Reinhardt (2008) suggest export variance as one of the possible

measure of export volatility, so the variance as a measure of export variability used

in this research is consistent with the method used in the existing literature. Despite

that similarity, our main concern is the identification of a proper econometric model

with appropriate variables that can explain the difference in export variance across

countries.

Many classical and Bayesian model selection methods exist to address the

variable selection problems in economic analyses, and some of these models apply

sensitivity analysis in their approach. For example, Leamer (1985) proposed Extreme

Bound Analysis (EBA) that is expected to produce a consistent model whose

conclusions do not change even when the list of variables is slightly altered.

However, Sala-i-Martin (1997) criticizes the Extreme Bound analysis, as in the case of

EBA presenting a “nothing is robust” conclusion; he proposed examining the entire

distribution of the estimators and found that substantial numbers of variables are

robust in growth regressions. Another efficient method for model selection is the

48

Bayesian Averaging of Classical Estimates as proposed by Sala-i-Martin, Doppelhofer,

and Miller (2004), albeit eight of the eighteen regressors they found to be robust in

growth regression were later found to be sensitive when using a Robust Model

Averaging approach (Doppelhofer and Weeks, 2011).

In contrast to the sensitivity analysis, some other approaches use higher

association between two variables but a lower prediction error method to identify

the best predictors. These methodologies begin building a model with no variables,

but add variables based on their correlation to the response variable, which is what

LARS does. Efron et al. (2004) report that LARS is an improved version of previously

used Forward Selection and Forward Stage-wise Regression methods because it is

less greedy and more efficient. This method has been used in various studies within

and beyond economics. For example, McCann and Welsch (2007) use the method to

study the relation between air pollution and mortality in US metropolitan cities;

Khan, Van-Aelst and Zamar (2007) identify the predictors for murder rates across US

States; Angostinelli and Salibian-Barrera (2010) apply the LARS algorithm based on S-

estimators for regressions. All of these studies report one thing in common: LARS

methodology produces computationally efficient and robust results even in the

presence of unreliable data due to outliers. Therefore, by using LARS in AfT

effectiveness, we expect a parsimonious set of explanatory variables for the efficient

prediction of export variability. Details of the LARS algorithm are explained in the

section that follows.

49

4.3. Data and Model

This study utilizes annual level longitudinal export data of AfT-recipient

developing countries over a period of sixteen years, from 1995 to 2010. Because we

are mainly interested in export variability, we calculated sample variance of exports

during that period using the equation below.

𝐸𝑥𝑝𝑜𝑟𝑡 𝑉𝑎𝑟𝑖𝑎𝑛𝑐𝑒: 𝜎𝑖2 = 1 𝑇

∑ (𝑋𝑖𝑡 − µ𝑖)216𝑡=1 (4.1)

Where, µ𝑖 = 1 𝑇

∑ (𝑋𝑖𝑡)16𝑡=1

Where, X denotes the export volume measured in US Dollar, ‘i’ represents

121 AfT-receiving developing countries, and ‘t’ represents time period. Our data are,

therefore, purely cross-sectional for this purpose, and comprises 121 countries in

the sample. Although AfT was designed primarily to help the least developed

countries, there are a number of middle-income countries benefitting as well.16 As

such, this study considers countries categorized as low-income and middle-income

by the World Bank in its analysis.

The export data are obtained from United Nation’s Conference on Trade and

Development (UNCTAD), except for the service exports which are obtained from the

World Development Indicator database maintained by the World Bank. The aid data

are obtained from OECD’s Creditor Reporting System (CRS). Other control variables

included in the analysis are GDP per-capita, financial development (using bank

private-credit to GDP ratio as a proxy), exchange rate, trade openness (using trade

16 See Appendix A for list of countries included in the study.

50

freedom as a proxy), and institutional quality (using regulatory quality as a proxy).

GDP per-capita and exchange rate are obtained from the World Development

Indicator database; measure of financial development is obtained from the Global

Financial Development database; trade openness is obtained from the Heritage

Foundation; and, finally, the institutional variables are obtained from the World

Governance Indicator database maintained by the World Bank.

Since total exports from a country are likely to depend on a number of

macroeconomic variables as well as the presence of quality institutions and their

policies towards the trade sector, Aid-for-Trade is interacted with these macro and

institutional variables. Mainly we are interested in the interaction of foreign-aid with

regulatory quality (RQ), trade openness (TO), and financial development (FD). It is

well established by Burnside and Dollar (2000) that the aforementioned variables

are important determinants of the effectiveness of foreign aid in general. In

addition, GDP per-capita and exchange rates (EXR) are also important variables in

export analysis. Hence, we also include their interaction with AfT. Because we are

trying to determine what variables should be included in the model, we incorporate

all potentially relevant variables.

Once we have the above set of possible independent variables, our concern

is the conformity of left-hand-side and right-hand-side variables in a regression

analysis. When we analyze export levels, it is plausible to use the set of covariates in

their level form. However, the analysis of export variance with respect to variance of

51

the covariates might have better predictive power. Our goal is to select the

consistently significant set of variables from the pool of variables in their level form,

variance form, and a combination of interaction terms in analyzing the export

variance.

In order to accomplish the task of proper model selection, we use the Least

Angle Regression (LARS) approach as proposed by Efron et al. (2004). LARS is a

newer approach used in this type of model selection problem and the authors view

it as a more ‘democratic’ version of the Forward Stepwise Regression. Hastie et al.

(2009) have shown that LARS is very closely related with Least Absolute Shrinkage

and Selection Operator (LASSO) and, in fact, provides an efficient algorithm for

running the entire LASSO path. Therefore, we focus on LARS algorithm in this

analysis. LARS builds a model adding “as much” of a predictor in succession as the

model deserves based on the correlation between the dependent variable and the

independent variable. In our examination with LARS, the first step involves the

identification of a variable most correlated with export variance. However, LARS

does not completely fit the variable in the model, but rather moves the coefficient of

this variable continuously toward its least square value. In doing so, the correlation

between the coefficient and residual from that particular model decreases in

absolute value. Then, as soon as another variable “catches up” in terms of

52

correlation with the residual, the process is paused and the second variable joins the

active set. This process is continued until all the variables are in the model.17

Since our dependent variable is a sample variance, we would like to know

which independent variables conform to the dependent variable. So, the two main

questions we are trying to answer are: first, whether we need to use the right-hand-

side variables in levels or variance form; and second, if AfT consistently stands out as

one of the significant variables. With this in mind, we establish two of the largest

models that can be constructed using a combination of the explanatory variables,

where export variance is a function of all other variables. First, we start building a

model that includes explanatory variables in level as well as variance form and their

interaction with AfT and AfT-variance, as shown in Model 1. The acronyms have

their corresponding meaning as explained at the beginning of this section, a ‘V’ in

front of the acronym means the variance, and ‘*’ denotes the interaction between

two variables.

Model 1:

𝐸𝑥𝑝𝑜𝑟𝑡 𝑉𝑎𝑟𝑖𝑎𝑛𝑐𝑒 = 𝑓(𝐴𝑓𝑇,𝐺𝐷𝑃,𝐸𝑋𝑅,𝐹𝐷,𝑇𝑂,𝑅𝑄,𝑉𝐴𝑓𝑇,𝑉𝐷𝐺𝑃,𝑉𝐸𝑋𝑅,𝐴𝑓𝑇 ∗ 𝑉𝐴𝑓𝑇,𝐴𝑓𝑇 ∗ 𝐺𝐷𝑃, 𝐴𝑓𝑇 ∗ 𝐸𝑋𝑅, 𝐴𝑓𝑇 ∗ 𝑅𝑄, 𝐴𝑓𝑇 ∗ 𝑇𝐹, 𝐴𝑓𝑇 ∗ 𝐹𝐷, 𝑉𝐴𝑓𝑇 ∗ 𝐸𝑋𝑅,𝑉𝐴𝑓𝑇 ∗ 𝑅𝑄, 𝑉𝐴𝑓𝑇 ∗ 𝑇𝑂, 𝑉𝐴𝑓𝑇 ∗ 𝐹𝐷)

Next, we build an even more comprehensive model by including interaction

of AfT and AfT-variance with other control variables in their variance form. Doing so

gives us an opportunity to select the most significant variables from the pool of all

17 A step-by-step algorithm for LARS/LASSO can be found in Hastie et al. (2009).

53

possible regressors that are expected to influence export variability. The exact sets

of variables in the second model are identified below.

Model 2:

𝐸𝑥𝑝𝑜𝑟𝑡 𝑉𝑎𝑟𝑖𝑎𝑛𝑐𝑒 = 𝑓(𝐴𝑓𝑇,𝐺𝐷𝑃,𝐸𝑋𝑅,𝐹𝐷,𝑇𝑂,𝑅𝑄,𝑉𝐴𝑓𝑇,𝑉𝐷𝐺𝑃,𝑉𝐸𝑋𝑅, 𝐴𝑓𝑇 ∗ 𝑉𝐴𝑓𝑇, 𝐴𝑓𝑇 ∗ 𝐺𝐷𝑃, 𝐴𝑓𝑇 ∗ 𝐸𝑋𝑅, 𝐴𝑓𝑇 ∗ 𝑅𝑄, 𝐴𝑓𝑇 ∗ 𝑇𝑂, 𝐴𝑓𝑇 ∗ 𝐹𝐷, 𝑉𝐴𝑓𝑇 ∗ 𝐸𝑋𝑅,𝑉𝐴𝑓𝑇 ∗ 𝑅𝑄, 𝑉𝐴𝑓𝑇 ∗ 𝑇𝑂, 𝑉𝐴𝑓𝑇 ∗ 𝐹𝐷, 𝐴𝑓𝑇 ∗ 𝑉𝐺𝐷𝑃,𝐴𝑓𝑇 ∗ 𝑉𝐸𝑋𝑅, 𝑉𝐴𝑓𝑇 ∗ 𝑉𝐺𝐷𝑃, 𝑉𝐴𝑓𝑇 ∗ 𝑉𝐸𝑋𝑅 )

Once all of the variables are included in the model, the determination of

significant variables is based on Mallows’s Cp following Efron et al. (2004)18. As we

can see in the results, the Cp statistic decreases as we increase the set of active

variables up to a certain point. After that, the statistic increases as the correlation

between the new variable and the residual decreases. Based on this information, we

select the set of variables that produce the lowest Cp statistic. These variables would

be the basis for our analysis by which we examine the nuances of the relationship

between export variance and AfT.

4.4. Empirical Results

This section explains the results obtained from empirical analysis of the

relationship between export variance and the set of possible explanatory variables.

We present the results for two of the biggest models that can be constructed based

on the combinations of independent variables. In the first table each set of results

shows the LARS algorithm, where variables are added to the model based on their

18 The Cp criterion was introduced by Mallow (1973) to use with Ordinary Least Square as an unbiased estimator for the model error.

54

correlation to the dependent variable, with the first variable on the list having the

highest correlation followed by the one with a lower correlation at each step. As

explained in the algorithm for LARS method, our goal here is to find a model with

the lowest Cp statistic, as indicated by ‘*’ after the Cp value in the first table. The

second table shows, in each set of results, the coefficients for each independent

variable whose combination produces a model that has the smallest Cp statistic.

We are mainly interested in the role of AfT in reducing export variability to

examine if such a relation exists in our sample of developing countries. The major

concern in our analysis is the use of a proper combination of independent variables,

and specifically, if AfT has a significant impact on export variability in either form.

Table 8 presents the results obtained from LARS analysis of Model 1. We see

that GDP-variance and AfT-variance interacted with regulatory quality seem to

explain most of the export variance from developing countries. We see that higher

variance in per-capita GDP results in higher export variance. However, even if AfT is

associated with higher variance, those countries with better regulatory quality can

reduce export variance. Remarkably, neither AfT nor other variables seem to have

significant impact on export variance in their level form.

Moving onto Table 9, the results obtained by expanding the model by adding

the interaction of AfT and AfT-variance with the variance of other regressors. We see

similar results with respect to GDP-variance and the interaction between AfT-

variance and regulatory quality as in the first model. However we see, additionally,

55

that if AfT is allocated with a greater variance to those countries which have greater

GDP-variance, there is an increase in export variance. This is shown by a positive

coefficient for the interaction between AfT-variance and GDP-variance.

In answer to our inquiry, the results show that the regressors in variance

form have the best predictive power when we are explaining a dependent variable

in its variance form. In particular, although we do not see a significant impact of AfT

or AfT-variance on its own, we see that smooth allocation of AfT to countries with

lower GDP-variance has more favorable effects in general. This claim is supported by

a positive coefficient for the interaction between AfT-variance and GDP-variance.

One implication of this finding is that AfT donors should be considering the

economic policies in aid receiving countries because it is generally understood that

low GDP-variance is likely the result of good economic policies. Another implication

is that donors should distribute aid carefully and consistently because higher AfT-

variance may indicate poor channeling of resources and a lack of investment in trade

promoting sectors.

In addition, we also see that AfT when interacted with regulatory quality

turns out to be one of the significant variables in explaining export variability in the

sample, suggesting that better regulatory quality augments the favorable effects of

AfT. This finding is somewhat similar to the argument that foreign aid is more

effective in a good policy environment. In creating pro-business regulations, i.e.

presence of better regulatory quality, a government supports a favorable business

56

environment which may be an indication that other sound economic policies are

already in place in the aid recipient country. Therefore, in distributing aid, donors

may want to consider closely the assurance of an environment for private sector

development. These findings are somewhat consistent with the literature that

suggests foreign aid is effective in a good policy environment.

4.5. Conclusions with Discussions

This chapter contributes to the discussion of foreign-aid effectiveness by

addressing two major concerns: first, conformity of explanatory variables when a

model consists of variance as the response variable; and, the second, if AfT, in one

form or other, helps predict variability of exports in the AfT recipient countries. In

answer to our first query, we find that AfT-variance, GDP-variance, and regulatory

quality consistently show the most predictive power for export variance of

developing countries. It addresses the variable selection problem in the analysis of

foreign aid effectiveness; it suggests that the explanatory variable should be used in

their variance form when the response variable is chosen to be in variance form. The

results indicate that the majority of variance in exports can be predicted by GDP-

variance. Assuming that economic policies dictate GDP-variance, the role of national

governments, private sectors, and international communities are equally important

in implementing better economic policies. This would ultimately help gain a

sustained export industry in developing countries.

57

Additionally, based on the association and reduced prediction error, AfT-

variance turns out to be an important regressor to explain export variance. We see

that higher AfT-variance is detrimental to export performance of a developing

country when the country is experiencing higher GDP-variance, since that would

result in higher export variance. However, AfT can reduce the variability of exports in

the presence of better regulatory quality even if AfT is allocated with a higher

variance, suggesting an important role of institutional policy in the reduction of

export variance, and thus supporting the argument that foreign aid is more effective

in a good regulatory environment, somewhat akin to the good policy argument in

aid and growth literature. This research finds that AfT actually produces favorable

results in the exporting sectors of the aid recipients. Such results call for a smooth

allocation for AfT for trade related activities from donors’ side while it highlights the

role of governments and private sectors to come up with good economic policies.

The overall results of the relationship between export variance and AfT-

variance may prove valuable in the future when researchers, aid donors, and policy

makers want to consider the effectiveness of financial assistance to developing

countries. Conceptually, AfT is presumably more effective when it raises export

levels but also enables a low export variance. From a policy standpoint, favorable

results in terms of both the level of exports and its variance call for increased

allocation of aid that is targeted to exports. Finally, from a development perspective,

AfT’s positive role in enhancing and maintaining higher exports indicates better

58

economic prospects and likely the beginning of sustained and balanced economic

growth.

59

CHAPTER 5

CONCLUSIONS

Empirical research has not shown conclusively that foreign aid has a

significant impact on overall economic growth. Previous studies debated for or

against foreign aid by examining the effects on overall economic growth due to

overall Official Development Assistance (ODA); this produced results that showed

inconclusive aggregate growth effects but did not address specific impacts of

targeted aid. This dissertation argues that a different approach to measuring foreign

aid, one that is focused on achieving a specific economic outcome, may prove

valuable for evaluating the benefits of foreign aid. Concentrating on a particular

economic outcome with respect to foreign aid reveals its truer impact. Accordingly,

this research demonstrates that an analysis of export performance with respect to

foreign aid that is exclusively targeted for trade sector improvement (Aid-for-Trade

or AfT) produces favorable results. This has significant consequences for the broad

argument that aid is valuable and effective for developing countries, because export

growth ultimately influences economic growth and development. Therefore, the

policy prescription is that effectiveness of foreign aid should be evaluated on the

basis of the immediate objectives of aid and whether or not those goals are met.

The results in this dissertation are derived from three separate essays that

each use export data of 121 AfT-recipient countries over a period of 16 years [1995-

60

2010] to examine the role of AfT on export performance. The three studies find in

favor of AfT, specifically in terms of the effects on aggregate export levels,

disaggregated sector specific export levels, and variability of aggregate exports.

Regardless of the particular aspect of exports examined, the results confirm that

export-targeted aid is successful in meeting the main objectives.

The first essay contributes to the literature by scrutinizing the effectiveness

of foreign aid on export performance of aid recipient countries in a comprehensive

way. The impact of Aid-for-Trade on export performance is analyzed at the

aggregate level of exports considering various categories of AfT based on their origin

(bilateral AfT vs. multilateral AfT) as well as end use (narrower vs. broader measure

of AfT). Overall results show that total exports from aid-recipient countries

significantly increase as a result of AfT. By using multiple disaggregated measures for

AfT and consistently finding favorable impact on export performance, our results

provide a stronger support towards the effectiveness of such targeted aid; this

showcases why it is sensible for donors to provide additional AfT and to positively

affect the economic performance of developing countries. The aggregate analysis

also examines the impact of aid on annual export growth pattern, as well as change

in the share of export to GDP. Both of these measures exhibit a positive and

significant impact as a result of increased AfT.

Favorable gains in aggregate exports as a result of AfT are laudable provided

some key sectors are not negatively impacted in the process. Using a stringent

61

criterion, AfT cannot be effective if there is no positive gain in each of the

disaggregated sectors. Therefore, the second essay contributes to the study of aid

effectiveness by analyzing the impact of sectoral aid on corresponding sector-

specific exports using the Seemingly Unrelated Regression (SUR) technique; this

approach produces efficient estimates after taking care of inter-sector

interdependence. The study constructs sectoral aid measures for agriculture,

manufacturing, and service sectors, which are then seen to exert a positive and

statistically significant influence on sectoral exports in the corresponding export

sectors. Results presented are robust to different model specification and various

measures of sectoral-aid and suggest a favorable role for aid that is specifically

targeted to sector-specific exports.

While the first two essays are concerned about increasing the total volume of

exports from aid recipients, the third essay is concerned about the consistency of

the positive results obtained there. The encouraging results in the previous two

cases would not be adequate if they are achieved at the cost of increased export

variability. Fortunately, the results indicate that AfT does not only increase export

levels but also helps reduce variability of exports. More importantly, the chapter

contributes to the discussion of foreign-aid effectiveness by addressing the model

selection problem and identifying AfT as an important variable to improve export

performance of developing countries. We find that AfT-variance as well as GDP-

variance has the most predictive power for export variance of developing countries.

62

This addresses the variable selection problem in the analysis of foreign aid

effectiveness, suggesting in this case the use of variance-form regressors. The policy

implications of these findings are twofold: (a) as lower export variance is expected to

depend on lower AfT-variance, donors should attempt to be consistent in allocating

such aid, and (b) as GDP-variance explains majority of export variance, aid recipients

should concentrate on adopting economic policies that are conducive to steady

economic performance. Additionally, our finding that shows a significant role of AfT

in reducing export variance in the presence of better regulatory quality supports the

argument that foreign aid is effective in better policy environments. This calls for

allocation of foreign aid conditional on commitments to improving economic policies

in developing countries.

Overall, the study finds that AfT has a positive impact on each of the

different measures of export performance considered in this dissertation; this

further calls for increased export-targeted foreign-aid to developing countries since

many countries still have yet to realize the widespread benefits that could be

achieved with improved employment and incomes that come with a competitive

trade sector. The policy issue donors should consider is to replace ‘general’

budgetary support with specific ‘targeted’ aid. The disappointing results seen in

growth regressions may be attributable to the donors not making careful decisions

about where or how to disburse their foreign aid. The empirical results show that

targeted aid produces favorable outcomes and suggests that donor countries focus

63

on providing sector specific aid if they truly want to help the development of aid

recipient countries. Conceptually, these findings are encouraging because of AfT’s

positive role in increasing the level of exports while reducing its variability, both of

these measures being important for a healthy economic performance of any

country. From a policy standpoint, it makes perfect sense to emphasize a specific

category of foreign aid that produces positive results at a time when the

effectiveness of general aid is questionable. These results are also in agreement with

the development perspective: AfT’s positive role in enhancing and maintaining

higher exports indicates better economic prospects and likely sustained and

balanced future economic growth.

Despite these promising claims, out study also suggests the need for more

research in this particular area to highlight the connections between sectoral aid and

exports during the different stages of economic development. To the extent that

trade happens to be a prominent engine of growth and development for many

recently emerging countries, it would help to identify precisely when the potential

effect of aid on trade is sizeable. For example, for aid resources to be appropriately

employed, it would be useful to know if aid-for-trade is more effective in low-

income or middle-income countries, so that aid resources could be appropriately

employed.

Given that these results are based on the average impact of aid on all AfT

recipients, individual examinations may reveal a more detailed picture. Case studies

64

of individual countries would likely provide particulars of the realities facing these

countries, including the limits imposed by diminishing returns. Future work must be

geared toward such an approach. Improved export performance of developing

countries depends on many more factors than the financing of projects by foreign

aid; effectiveness of foreign aid will be significant when a proper economic

environment for promoting exports is in place in developing countries for the long

run. An interesting extension to this research would be to consider: (a) examining

the impact of AfT on the productivity of exporting sectors (export quality), and (b)

exploring the connections between AfT, the political economy of trade, and

economic development. The rich dataset put together in this research, both sector

level disaggregated AfT and sectoral exports, could be used to capture key relations

in this particular area by applying additional econometric modeling techniques. This

dissertation in conjunction with such studies could be important to the

understanding of the political economy of foreign aid and international trade and

their subsequent effects on economic development.

65

REFERENCES

1. Agostinelli, C. and M. Salibian-Barrera (2010). Robust Model Selection with LARS

Based on S-estimators. Proceedings of COMPSTAT’ 2010. 19th International

Conference on Computational Statistics. Paris, France. August 22 – 27, 2010.

2. Alesina, A. & Dollar, D. (2000). Who Gives Foreign Aid to Who and Why? Journal

of Economic Growth. 5:1. 33 – 63.

3. Arellano, C., A. Bulir, T. Lane, & L. Lipschitz (2009). The Dynamic Implications of

Foreign Aid and its Variability.

4. Arellano, M. & S. Bond. (1991) Some Tests of Specification for Panel Data: Monte

Carlo Evidence and Application to Employment Equations. The Review of

Economic Studies. 58: 277 – 297.

5. Arellano, M. and O. Bover. (1995). Another Look at the Instrumental Variable

Estimation of Error Components Models. Journal of Econometrics. 68:1. 29-51.

6. Awokuse, T. (2006). Causality between Exports, Imports, and Economic Growth:

Evidence from Transition Economics. Economic Letters. 94(3): 389-395.

7. Balassa, B. (1978). Exports and Economic Growth: Further Evidence. Journal of

Development Economics. 5(2): 181-189.

8. Beck, T. (2002). Financial Development and International Trade – Is there a Link?

Journal of International Economics. 57: 107 – 131.

9. Bhagwati, J. (1958). Immiserizing Growth: A Geometrical Note. The Review of

Economic Studies. 25:3. 201 – 205.

10. Brazys, S. (2010). Race to Give? The Selective Effectiveness of United States

Trade Capacity Building Assistance. Review of International Political

Economy. 17:3, 537 – 561.

11. Bruesch, T. & A. Pagan (1980). The Lagrange Multiplier Test and its Application to Model

Specification in Econometrics. The Review of Economic Studies. 47(1): 239-253.

12. Burnside C, Dollar D. 2004. Aid, Policies and Growth: Revisiting the Evidence.

World Bank Policy Research Paper. No. O-2834. The World Bank: Washington DC.

66

13. Burnside, C. & Dollar, D. (2000). Aid, Policies, and Growth. American Economic

Review. 4 (90): 847–868.

14. Cali, M., M. Razzaque & D. te Velde (2011). Effectiveness of Aid for Trade in Small

and Vulnerable Economies. Economic Paper. 91.

15. Chong, A., M. Gradstein, C. Calderon (2009). Can Foreign Aid Reduce Income

Inequality and Poverty? Public Choice. 140: 59 – 84.

16. Collier P, Dehn J. 2001. Aid, Shocks and Growth. Policy Research Working Paper

No.2688. The World Bank: Washington DC.

17. Collier P, Dollar D. 2002. Aid allocation and poverty reduction. European

Economic Review. 46(8): 1475–1500.

18. Dalgaard CJ, Hansen H. 2001. On aid, growth and good policies. Journal of

Development Studies. 37(6): 17–41.

19. Doppelhofer, G. and M. Weeks (2011). Robust Growth Determinants. Institutt for

Samfunnsokonomi. ISSN: 0804 –6824.

20. Doucouliagos, H. and Paldam, M. (2009). The Aid Effectiveness Literature: The

Sad Results of 40 Years of Research. Journal of Economic Surveys. 23 (3) : 433 –

461.

21. Easterly W., R. Levine, and D. Roodman (2004). New data, new doubts: a

comment on Burnside and Dollar’s Aid, policies and growth (2000). American

Economic Review 94(3): 781–784.

22. Easterly, W. & A. Kraay (2000). Small States, Small Problems? Income, Growth

and Volatility in Small States. World Development. 28(11): 2012-2017.

23. Easterly, W. (2006). The White Man’s Burden: Why the West’s Efforts to Aid the

Rest Have Done So Much ill and So Little Good. New York: Penguin Press HC

24. Efron, B., T. Hastie, I. Johnstone & R. Tibshirani (2004). Least Angle Regression. Annals of

Statistics. 32: 2. 407 – 499.

25. Feder, G. (1983). On Exports and Economic Growth. . Journal of Development

Economics. 12(1-2): 59-73.

67

26. Global Financial Development (2011). The World Bank. Washington, D.C.

27. Haddad, M., J. Lim & C Saborowski (2009). Trade Openness Reduces Growth

Volatility When Countries are Well Diversified. Working.

28. Hansen, H. and F. Tarp (2001). Aid and Growth Regressions. Journal of Development

Economics. 64: 547 – 570.

29. Hastie, T., R. Tibshirani & J. Friedman (2009). The Elements of Statistical Learning: Data

Mining, Inference, and Prediction. Springer.

30. Hausmann, R. & Rodrik, D. (2006). What You Export Matters. Journal of Economic

Growth. 12(1): 1-25.

31. Helpman, E., M. Melitz, and S. Yeaple (2004). Export Versus FDI with Heterogeneous

Firms. The American Economic Review. 94: 300 – 316.

32. Heritage Foundation (2012). Economic Freedom Index. http://www.heritage.org/index/

. Accessed May 2012.

33. Islam N. 2005. Regime Changes, Economic Policies and the Effect of Aid on

Growth. The Journal of Development Studies. 41(8): 1467 – 1492.

34. Kaufmann, D., A. Kraay, and M. Mastruzzi (2010). "The Worldwide Governance

Indicators: Methodology and Analytical Issues". World Bank Policy Research

Working Paper. No. 5430.

35. Khan, J., Van-Aelst, S. and R. Zamar. (2007). Robust Linear Model Selection Based

on Least Angle Regression. Journal of the American Statistical Association.

102(480): 1289 – 1299.

36. Krueger, A. (1998). Why Trade Liberalization is Good for Growth? The Economic

Journal. 108(450): 1513 – 1522.

37. Leamer, E. (1985). Sensitivity Analysis Would Help. The American Economic

Review. 75(3): 308 – 313.

38. Lloyd, T., McGillivray, M., Morrissey, O. & Osei, R. (2000). ‘Does aid create

trade?. European Journal of Development Research. 12(1): 107-123

39. Lloyd, T., McGillivray, M., Morrissey, O. & Osei, R. (2000). ‘Does aid create trade?.

European Journal of Development Research. 12(1): 107-123

68

40. Mallows, C. (1973). Some Comments on Cp. Technometrics. 15: 661 – 675.

41. Mansfield, E. & Reinhardt, E. (2008). International Institutions and the Volatility of

International Trade. International Organization. 62: Fall 2008. 621 – 52.

42. McCann, L. and R. Welsch (2007). Robust Variable Selection Using Least Angle

Regression and Elemental Set Sampling. Computational Statistics and Data Analysis. 52:

249 – 257.

43. McGillivray, M., S. Feeny, N. Hermes, and R. Lensink (2006). Controversies Over

the impact of Development Aid: It Works; It Doesn’t; It Can, But That Depends.

Journal of International Development. 18: 1031 – 1050

44. Melitz, M. J. 2003. "The Impact of Trade on Intra-Industry Reallocations and

Aggregate Industry Productivity," Econometrica, 71(6).

45. Morrissey, O. (2006). Aid or Trade, or Aid and Trade? The Australian Economic

Review. 39(1): 78–88.

46. Moyo, D. (2009). Dead Aid: Why Aid Is Not Working and How There Is Another

Way for Africa. New York: Farrar, Straus, and Giroux

47. Munemo, J. (2011). Foreign Aid and Export Diversification in Developing Countries. The

Journal of International Trade and Economic Development. 20: 3. 339 – 355.

48. OECD/WTO (2011), Aid for Trade at a Glance 2011: Showing Results.

http://dx.doi.org/10.1787/9789264117471-en

49. Organization for Economic Co-operation and Development (2006). Aid for Trade:

Making it Effective. The Development Dimension.

50. Organization for Economic Co-operation and Development (2012). Creditor

Reporting System. http://stats.oecd.org/Index.aspx?DatasetCode=CRSNEW .

Accessed May 2012.

51. Pettersson, J. & Johansson, L. (2011). Aid, Aid for Trade, and Bilateral Trade: And

Empirical Study. The Journal of International Trade and Economic Development.

52. Rajan, R. & Subramanian, A. (2008). Aid and Growth: What does the Cross

Country Evidence Really Show? Journal of Economics and Statistics. 90(4): 643-

665.

69

53. Sachs J., J. McArthur, G. Schmidt-Traub, M. Kruk, C. Bahadur, M. Faye, and G.

McCord (2004). Ending Africa’s poverty trap. Brookings Papers on Economic

Activity. 1:117–216

54. Sala-i-Martin, X. (1997). I Just Ran Four Million Regressions. NBER Working Paper

Series # 6252.

55. Sala-i-Martin, X., G. Doppelhofer, and R. Miller (2004). Determinants of Long-

Term Growth: A Bayesian Averaging of Classical Estimates (BACE) Approach. The

American Economic Review. 94 (4): 813 – 835.

56. StataCorp. 2011. Statastical Software: Release 11.0. College Station. TX: Stata

Corporation.

57. Svenssen, J. (2000). When is Foreign Aid Policy Credible? Aid Dependence and

Conditionality. Journal of Development Economics. 61(1): 61-84.

58. United Nations Conference on Trade and Development (UNCTAD) (2012).

UNCTADSTAT. http://unctadstat.unctad.org/ReportFolders/reportFolders.aspx?sCS_refe

rer=&sCS_ChosenLang=en . Accessed May 2012.

59. Wagner, D. (2003). Aid and Trade – An Empirical Study. Journal of the Japanese

and International Economics. 17: 153-173.

60. Winters (2010). Choosing to Target What Types of Countries Get Different Types

of World Bank Projects. World Politics. 62: 3. 422–58.

61. World Bank, 1998, Assessing Aid: What Works, What Doesn’t, and Why. World

Bank Policy Research Report. Oxford: Oxford University Press.

62. World Development Indicators (2011). The World Bank. Washington, D.C.

http://data.worldbank.org/data-catalog/world-development-indicators

63. World Trade Organization (2006). Recommendations of the Task Force on Aid-for-Trade.

WT/AFT/1 – 06-3617.

64. Worldwide Governance Indicators (2011). The World Bank. Washington, D.C.

http://databank.worldbank.org/data/views/variableselection/selectvariables.aspx?sour

ce=worldwide-governance-indicators

70

65. Zellner, A. (1962). An Efficient Method of Estimating Seemingly Unrelated Regressions

and Tests for Aggregation Bias. Journal of the American Statistical Association. 57 (98):

348 - 368.

71

Appendix A: Official Development Assistance Foreign Aid Categories (Official Development Assistance) Not included in AfT Included in AfT

Education Category 1 Health Trade Policies & Regulations Food Aid Population Policies Category 2 Water and Sanitation Transport & Storage Government and Civil Society Communications Conflict Resolution and Peace Energy Social Infrastructure Environmental Protection Category 3 General Budget Support Banking & Financial Services Debt Relief Business & Other Services Humanitarian Aid Agriculture Emergency Relief Forestry Disaster Prevention Fishing Administrative Costs Industry Support to NGOs Mineral Resources & Mining Refugee Management Tourism

AfT as a Percent of ODA for Select Years 2002 2006 2010

AfT/ODA 24% 18% 29%

AfT Categories as a percent of total AfT Category 1 4% 4% 3% Category 2 52% 52% 60% Category 3 44% 44% 37%

Broad and Narrow Measures of Aid: AfT1 = Category 1 AfT 2 = Category 1 + Category 2 Total AfT = Category 1 + Category 2 + Category 3

Source: OECD

72

Appendix B: Summary Statistics (Logged Real Values)

Variable Mean SD Min Max

Total Export 21.71098 2.306137 15.27629 33.27481 Manufacturing 20.17389 2.838401 12.6005 28.32308 Agriculture 19.50341 2.376122 12.81487 28.65891 Service 20.18481 1.78941 15.04667 24.9679 Mining 19.77319 2.793515 11.33546 28.58641

Export Growth 0.0018119 0.2942804 -3.41573 2.391069 Share Change 0.0018304 0.0590197 -0.3885911 0.4082006 Export Variance 0.3686443 1.219514 0.0023878 10.77033

Aid Variance 1.245166 2.736098 0.0126745 25.65785 AFT1 13.03644 2.435642 2.98017 20.49552 AfT2 16.76738 2.680089 7.942404 26.6331 AfT 18.11981 1.849304 11.62564 28.4408 Bi-AfT 17.17084 2.17415 8.268145 27.20644 Multi-AfT 16.56545 2.632257 2.230593 28.09686

Income 7.325953 1.252625 3.592018 17.27936 Openness 61.49815 16.95288 0 95 Exchange Rate 4.135976 2.537803 0.3551887 10.81656 Money Supply 32.57938 3.213639 24.61211 41.97512 Corruption Control -0.4997098 0.6381646 -2.489213 1.563225 Regulatory Quality -0.5022135 0.7356837 -2.481155 1.587131

Number of Aid Recipients 121

73

Appendix C: List of Countries Included in the Study Afghanistan Albania Algeria Angola Argentina Armenia Azerbaijan Bangladesh Belarus Belize Benin Bhutan Bolivia Bosnia and Herzegovina Botswana Brazil Burkina Faso Burundi Cambodia Cameroon Cape Verde Central African Republic Chad Chile China Colombia Comoros Congo, Dem. Rep. Congo, Rep. Costa Rica Cote d'Ivoire Djibouti Dominica Dominican Republic Ecuador Egypt, Arab Rep. El Salvador Ethiopia Fiji

Gabon Gambia, The Georgia Ghana Grenada Guatemala Guinea Guinea-Bissau Guyana Haiti Honduras India Indonesia Iran, Islamic Rep. Iraq Jamaica Jordan Kazakhstan Kenya Kosovo Kyrgyz Republic Lao PDR Lebanon Lesotho Liberia Macedonia, FYR Madagascar Malawi Malaysia Maldives Mali Mauritania Mauritius Mexico Moldova Mongolia Montenegro Morocco Mozambique

Namibia Nepal Nicaragua Niger Nigeria Pakistan Panama Papua New Guinea Paraguay Peru Philippines Rwanda Samoa Sao Tome and Principe Senegal Serbia Seychelles Sierra Leone Solomon Islands South Africa Sri Lanka St. Lucia St. Vincent and the Gren Sudan Suriname Swaziland Syrian Arab Republic Tajikistan Tanzania Thailand Timor-Leste Togo Tonga Tunisia Turkey Uganda Ukraine Uruguay Vanuatu

Venezuela, RB Vietnam Yemen, Rep. Zambia Zimbabwe

74

Appendix D

Supplemental Tables to Chapter 2

Aid-for-Trade and Aggregate Export Performance of Developing Countries: An Empirical Analysis

75

Table:1 Dependent Variable: Total Export. Estimation Method: System-GMM (1) (2) (3) (4) (5)

Income 0.5891*** 0.6542*** 1.0710*** 0.5512*** 1.0138*** (0.0280) (0.0312) (0.0311) (0.0291) (0.0316)

Openness -0.007904*** -0.007389*** -0.0003769 -0.005467*** -0.002743*** (0.000478) (0.000760) (0.000491) (0.000461) (0.000478)

Money Supply 0.1509*** -0.09723*** -0.1300*** 0.03312** 0.06341*** (0.0228) (0.0273) (0.0283) (0.0167) (0.0148)

Exchange Rate -0.03408 0.1049*** 0.3729*** 0.02271 0.2856*** (0.0272) (0.0246) (0.0280) (0.0208) (0.0246)

Corruption Control -0.2747*** -0.01664 -0.01737 -0.2206*** -0.1973*** (0.0253) (0.0260) (0.0334) (0.0185) (0.0217)

Regulation (RQ) 0.2400*** 0.1171*** -0.02817 0.1501*** -0.03125 (0.0454) (0.0349) (0.0221) (0.0464) (0.0347)

AfT1 -0.003006*** (0.000664)

AfT2 0.001931* (0.00111)

AfT 0.02761*** (0.00237)

Bi-AfT 0.009011*** (0.00316)

Multi-AfT 0.004931*** (0.000850)

N 600 600 600 600 600 P_Sargan 0.9996 0.9983 0.8936 0.9986 0.9998 P_AR2 0.4188 0.6712 0.3310 0.9762 0.8844

Standard errors in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01

76

Table:2 Dependent Variable: Total Export. Estimation Method: System-GMM

(1) Income 1.0217***

(0.0258) Openness -0.0004725

(0.000496) Money Supply -0.1312***

(0.0240) Exchange Rate 0.3362***

(0.0235) Corruption Control 0.01803

(0.0287) Regulation (RQ) 0.008331

(0.0238) AfT 0.1763***

(0.0483) AfT-Squared -0.004291***

(0.00139) N 600 P_Sargan 0.9502 P_AR2 0.3412 Standard errors in parentheses * p < 0.10, ** p < 0.05, *** p < 0.01

77

Table:3 Dependent Variable: Total Export. Estimation Method: System-GMM

(1) Income 1.0454***

(0.0321) Openness -0.0003835

(0.000352) Money Supply -0.1406***

(0.0243) Exchange Rate 0.3529***

(0.0295) Corruption Control -0.006493

(0.0303) Regulation (RQ) -0.007131

(0.0249) AfT 0.04771***

(0.00415) AfT-GDP Ratio -1.7847***

(0.392) N 600 P_Sargan 0.9177 P_AR2 0.3322 Standard errors in parentheses * p < 0.10, ** p < 0.05, *** p < 0.01

78

Table:4 Dependent Variable: Export Growth. Estimation Method: System-GMM

(1) (2) Income 0.5998***

(0.0339) Exchange Rate 0.2492***

(0.0316) Money Supply -0.1718***

(0.0271) Openness -0.002134*** -0.004696***

(0.000531) (0.000351) Regulation (RQ) -0.04650 0.1912***

(0.0296) (0.0252) Corruption Control -0.003048 -0.05518**

(0.0585) (0.0266) Income Growth 0.9342***

(0.0763) Exchange Rate Growth -0.04016

(0.109) Money Supply Growth 0.1466***

(0.0547) AfT 0.02539***

(0.00481) AfT Growth 0.01196***

(0.00220) N 415 415 P_Sargan 0.5072 0.9991 P_AR2 0.8244 0.2738 Standard errors in parentheses * p < 0.10, ** p < 0.05, *** p < 0.01

79

Table:5 Dependent Variable: Change in Share of Export to GDP. Estimation Method: System GMM

(1) (2) Income 0.04574***

(0.00753) Exchange Rate 0.05336***

(0.00659) Money Supply -0.03414***

(0.00235) Openness -0.0008015*** -0.001427***

(0.000163) (0.0000972) Regulation (RQ) 0.0005180 0.03848***

(0.00906) (0.00350) Corruption Control 0.04738*** 0.0004750

(0.00918) (0.00347) Income Growth 0.03250**

(0.0132) Exchange Rate Growth -0.05090***

(0.0154) Money Supply Growth 0.04264***

(0.00797) AfT 0.005256***

(0.000768) AfT Growth 0.007148***

(0.00142) N 414 414 P_Sargan 0.3039 0.9970 P_AR2 0.1671 0.2889 Standard errors in parentheses * p < 0.10, ** p < 0.05, *** p < 0.01

80

Appendix E

Supplemental Tables to Chapter 3

Sectoral Aid-for-Trade and Sectoral Exports: A Seemingly Unrelated Regression Analysis

81

Table:6 Dependent Variables: Sectoral Exports. Estimation Method: SUR. Cross-Sectional Study Narrowest Measure of SAfT Broadest Measure of SAfT

Agriculture Manufacturing

Service Agriculture Manufacturing

Service

Trade_Openness -0.0153 -0.00298 -0.0259** -0.0120 -0.00141 -0.0229** (0.0159) (0.0163) (0.0107) (0.0161) (0.0163) (0.0108)

Exchange_Rate -0.0305 -0.0155 -0.0282 -0.0326 -0.0217 -0.0402 (0.0644) (0.0668) (0.0442) (0.0652) (0.0666) (0.0446)

Corruption_Control -1.038** -1.124** -0.909*** -0.715 -0.856* -0.660* (0.482) (0.513) (0.337) (0.505) (0.517) (0.348)

Regulatory_Quality 1.849*** 1.014* 1.010*** 1.561*** 0.767 0.784** (0.494) (0.518) (0.339) (0.513) (0.521) (0.347)

Fin_Development 0.0149* 0.0145* 0.0195*** 0.0107 0.0133 0.0176*** (0.00820) (0.00866) (0.00571) (0.00834) (0.00863) (0.00578)

Agri_Value_Added 1.035*** 1.001*** (0.213) (0.208)

Man_Value_Added 0.911*** 0.924*** (0.123) (0.118)

Serv_Value_Added 0.582*** 0.621*** (0.0914) (0.0891)

A_AfT_Narrowest 0.472*** (0.0920)

M_AfT_Narrowest 0.492*** (0.0752)

S_AfT_Narrowest 0.367*** (0.0491)

A_AfT_Broadest 0.525*** (0.0944)

M_AfT_Braodest 0.601*** (0.0809)

S_AfT_Broadest 0.450*** (0.0549)

Observations 121 121 Standard errors in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01

82

Table:7 Dependent Variables: Sectoral Exports. Estimation Method: SUR. Panel Study Narrowest Measure of SAfT Broadest Measure of SAfT

Agriculture Manufacturing Service Agriculture Manufacturing Service Trade_Openness -0.0131*** -0.00380*** 0.00779*** 0.00385*** 0.00155 0.0206***

(0.00155) (0.00110) (0.00123) (0.00138) (0.00117) (0.00116) Exchange_Rate -0.203*** -0.108*** -0.164*** 0.0854*** 0.0547*** 0.0274***

(0.0126) (0.00902) (0.0101) (0.0105) (0.00858) (0.00898) Corruption_Control 1.101*** 1.178*** 0.369*** -0.800*** -1.273*** -0.699***

(0.0718) (0.0505) (0.0566) (0.0612) (0.0509) (0.0511) Regulatory_Quality -0.832*** -0.506*** -0.366*** 0.197*** -0.0172 -0.590***

(0.0709) (0.0528) (0.0570) (0.0582) (0.0524) (0.0497) Fin_Development -0.00297* -0.0192*** 0.0294*** 0.00218 -0.000673 0.0189***

(0.00157) (0.00111) (0.00125) (0.00137) (0.00114) (0.00104) Agri_Value_Added 1.135*** 1.162***

(0.0283) (0.0255) Man_Value_Added 0.827*** 1.111***

(0.0151) (0.0148) Serv_Value_Added 0.452*** 0.737***

(0.0186) (0.0175) A_AfT_Narrowest -0.0198**

(0.00811) M_AfT_Narrowest 0.159***

(0.00540) S_AfT_Narrowest 0.0950***

(0.00632) A_AfT_Broadest 0.170***

(0.00828) M_AfT_Braodest 0.281***

(0.00638) S_AfT_Broadest 0.233***

(0.00639) Observations 1936 1936

Standard errors in parentheses. * p < 0.10, ** p < 0.05, *** p < 0.01

83

Appendix F

Supplemental Tables to Chapter 4

Sectoral Aid-for-Trade and Sectoral Exports: A Seemingly Unrelated Regression Analysis

84

Table 8:

Model 1: Controls in levels + Controls in Variance + laft*vaft + laft*controls + vaft*controls

Step Cp R-square Action

1 404.4259 0.0000 +varpcap 2 24.6325 0.7407

3 5.2364 * 0.7822 +vaft_rq 4 6.7636 0.7832 +laft 5 7.3198 0.7860 +laft_lexr 6 8.1840 0.7882 +tfree 7 9.8612 0.7888 +vaft_priv2y 8 11.6912 0.7891 +laft_rq 9 13.6729 0.7892 +lpcap 10 15.5021 0.7895 +laft_priv2y 11 17.3909 0.7897 +vaft_lexr 12 16.8987 0.7945 +laft_lpcap 13 16.4525 0.7993 +varexr 14 10.2914 0.8151 +lexr 15 12.0095 0.8157 +priv2y 16 13.6031 0.8165 +vaft_laft 17 15.3435 0.8170 +rq 18 16.3016 0.8190 +laft_tfree 19 18.0787 0.8194 +varaft 20 20.0000 0.8196 +vaft_tfree

* indicates the smallest value for Cp

The coefficient values for the minimum Cp

Variable Coefficient

varpcap vaft_rq

0.7667 -0.0171

85

Table 9: Model 2: Controls in levels + Controls in Variance + laft*vaft + laft*controls + vaft*controls + laft*varcontrols + vaft*varcontrols

Step Cp R-square Action

1 389.1205 0.0000 +varpcap 2 93.0618 0.5960

3 6.6908 0.7727 +vaft varpcap 4 2.9366 * 0.7842 +vaft rq 5 4.8658 0.7843 +laft lexr 6 6.6938 0.7847 +lpcap 7 8.0273 0.7860 +laft varpcap 8 8.9622 0.7881 +laft 9 9.9309 0.7902 +tfree 10 11.1040 0.7918 +vaft varexr 11 12.0619 0.7939 +laft varexr 12 12.6997 0.7966 +laft lpcap 13 13.4521 0.7991 +vaft priv2y 14 15.0117 0.8000 +lexr 15 16.1163 0.8018 +laft priv2y 16 13.1318 0.8118 +rq 17 12.8680 0.8163 +vaft lexr 18 14.1460 0.8178 +priv2y 19 15.8703 0.8183 +laft rq 20 17.7991 0.8184 +laft tfree 21 19.6818 0.8187 +vaft tfree 22 20.6105 0.8208 +varexr 23 22.1593 0.8217 +varaft 24 24.0000 0.8220 +vaft_laft

* indicates the smallest value for Cp

The coefficient values for the minimum Cp

Variable Coefficient

varpcap vaft_rq vaft_varpcap

0.6423 -0.0063 0.0174

86